This chapter presents the results of our evidence review for the following four key questions (KQs): KQ 1, interaction of CHWs with participants; KQ 2, outcomes of community health worker (CHW) interventions; KQ 3, cost-effectiveness of CHW interventions; and KQ 4, training of CHWs and the relationship between CHW training and patient health outcomes. We note that KQ 3, on cost-effectiveness of CHW outcomes, is derivative of KQ 2 and is limited to studies demonstrating effectiveness. As noted in Chapter 2, a total of 53 studies qualified for inclusion for KQ 1 and KQ 2, 6 for KQ 3, and 9 for KQ 4.
Appendix C-1 provides the detailed evidence tables for KQs 1, 2, and 3. Appendixes C-2 and C-3 present individual quality ratings for randomized clinical trials (RCTs) and observational studies, respectively. Appendix C-3 provides detailed abstractions for KQ 4. All evidence tables are presented in alphabetical order by last name of the first author.
As noted in earlier chapters, an overall assessment of the effectiveness of CHW outcomes requires evaluation of sources of heterogeneity, including clinical context, intensity of interaction between CHWs and participants, and type of comparator. CHW interventions operate in a variety of clinical contexts; summarizing the effects of these interventions on varied outcomes requires an explicit consideration of the clinical context. For this reason, we have organized the results for KQs 1, 2, and 3 by the clinical context of the interventions identified. These are, specifically, health promotion and disease prevention, injury prevention, maternal and child health, cancer screening, and chronic disease management.
An additional source of heterogeneity is the degree of intensity of the intervention, which can vary by clinical context. We synthesize the evidence from KQ 1 to develop a measure of the intensity (low, moderate, or high) of the interaction between CHWs and participants, and we then include the measure in describing results for KQ 2 and KQ 3. We also record other sources of heterogeneity such as the type of comparator. Chapter 4 discusses the effectiveness of CHW interventions and the potential impact of sources of heterogeneity on effectiveness more fully.
This literature is characterized by several articles together constituting a single study. We refer to studies in the text and cite all relevant articles for each study; article and study counts, therefore, frequently do not match. Our summary tables below feature groups of studies addressing each outcome. Unless otherwise stated, these tables are organized alphabetically by the last name of the first author. The summary tables for KQ 2 and KQ 3 provide information to identify the study (author, and date of publication), study design, population and setting, sample size, study quality, intervention and comparators, and results.
KQ 1: Interaction of Community Health Workers and Participants
KQ 1 focuses on how CHWs interact with participants, specifically the place of service, type of service, type of educational materials used, duration of interaction with participants, and length of contact. We categorize place of service as over the telephone or based in the clinic, the community, home, or workplace. Interventions often employed multiple settings to interact with participants. The type of services ranged from one-on-one interactions to group interactions. All CHW interventions included some element of education; we sought to understand the degree to which these materials were standardized or tailored for each participant. We identified three elements of duration of interaction: the number of sessions, time per session, and the length of time from the first interaction to the last interaction (length of contact). We report summary findings below, for each descriptor, of the interaction between CHWs and participants across all studies and clinical contexts. These characteristics vary greatly across CHW interventions, but a common element is the overall intensity of the intervention. Interventions of lower intensity will require fewer resources than interventions of moderate or high intensity. As a proxy measure of resource allocation, we employ characteristics of the CHW-participant intervention to develop a measure of intensity of interaction. As noted earlier, a key organizing principle for understanding the effectiveness of CHW interventions is clinical context; we conclude this section by describing characteristics of CHW-participant interactions and their intensity by clinical context. Summary tables describing the characteristics of CHWs are provided by clinical context; within each table, studies are presented in order of intensity and then in alphabetical order, by the last name of the first author.
Overview of Interaction Between Community Health Workers and Participants
Place of service. CHWs interacted with participants over the telephone or provided services in one or more of four locations: home, community, clinic, or workplace. CHWs provided home, telephone, and clinic interventions on a one-on-one basis; community interventions were more likely to be oriented toward groups than to individuals. Thirty-two studies had at least one home visit but may have involved telephone and community components as well.15,17,18,23,59–102 Interventions in five studies occurred primarily by telephone.19–22,69,70,103–106 In nine studies, interventions included at least one meeting in a community setting and were primarily group-oriented.27,59,60,107–115 Interventions taking place in the community generally occurred in churches59,60,107,108 or in other neighborhood or community locations.27,102,109–114,116–122 Eight studies involved community interventions but did not specify the location;27,109–115,118–120 of these, only two were one-on-one interventions.118–121 One intervention occurred in a neighborhood beauty salon,116 and four occurred on the street or in shelters.117–122 Five studies took place within clinics or health care settings.23,77,99,123–125 One intervention occurred in the workplace.126 We could not determine the place of service for one intervention.16,127
Type of service. The type of services varied greatly across included studies. CHWs provided a wide range of services including one-on-one counseling (face-to-face and by telephone), education, support, information on health and community resources, transportation, appointment reminders, and other forms of assistance. The type of service ranged from brief one-time interactions to intensive one-on-one interactions over a span of years. The minimal service provided was a brief, one-time interaction such as distributing condoms and providing prevention literature117 or a single telephone call to promote cancer screening.103 At the other end of the spectrum, many interventions had multiple face-to-face counseling sessions, often in the home, to address specific needs.
Type of educational materials used. The least described characteristic of the interaction between CHWs and clients is the type of educational material used. As many as twenty-seven studies did not report any details on the type of educational materials utilized.16,19–23,67–74,78,79,83,85–92,96,97,99,101,104–108,117,120–122,124,125,127 Several studies did not describe educational materials per se but did report that they distributed “materials” as part of the intervention (e.g., safety glasses, materials to reduce exposure to asthma triggers, smoke detectors). The remainder provided minimal descriptions that ranged from the use of a postcard66 to complex systems, including audio and written formats to appeal to the broadest range of subjects.63
Duration of interaction (time per session and number of sessions). The duration of interaction varied broadly overall. Interactions lasted from quite brief (5 minutes to an hour) onetime meetings to extensive multiple interactions totaling several hours in all.
Length of contact. The length of contact—that is, the length of time that CHWs were directly involved with participants (which may have differed from the length of the study, or the length of time between measurement of pre- and postintervention health outcomes)—was inadequately reported in many cases. Length of contact ranged from 1 day15,103,108,116,117 to 2.5 years.98
Intensity of Interaction
Based on the type of interaction, the duration of interaction (time per session, number of sessions, and length of interaction), and the tailoring of CHW interactions, we classified the intensity of an intervention into three categories: low, moderate, or high. Interactions that had at least four of six elements suggesting a higher intensity (one-on-one, face-to-face, an hour per session or more, 3 or more months’ duration, three or more interactions, and tailored materials) were classified as high intensity. Interventions with two or three elements were classified as moderate intensity. Interventions with only one or none of the elements were classified as low intensity.
In making these classifications, we relied, whenever possible, on the protocol intentions rather than what actually occurred. When no information was available for the protocol, we relied on reported interactions in the field. When interactions in the field were also not reported, we assumed lower intensity for that aspect for the intervention. For instance, when studies did not report the time spent in each session, we assumed that the time per session did not exceed an hour on average. Similarly, if studies did not report specifically that the materials were tailored for each participant, we assumed that the interventions used generic materials for all participants.
Low-intensity interventions were generally one-time interactions, usually in a group setting. Moderate-intensity interventions occurred in a variety of settings but typically involved only one or two interactions with CHWs over shorter periods of time. High-intensity interventions included multiple interactions, face-to-face, for 3 months or more. Each category varies internally: for instance, within the high-intensity interventions, the number of interactions could vary from 3 to more than 20 in a year, depending on the nature of the intervention. Of the total of 53 studies, we classified 8 studies as low intensity,19–22,59,60,103,104,107,108,113,117,126 18 as moderate intensity106,125,15,23,63,66,69,70,99,101,102,105,109–112,114,116,118,119,122–124 and 27 as high intensity.16–18,27,61,62,64,65,67,68,71–98,100,120,121,127,128
Community Health Worker-Participant Interaction by Clinical Context
Community health worker-participant interactions for health promotion and disease prevention intervention. We included 11 studies on health promotion and disease prevention (Table 5). Six studies occurred in the home and by telephone;64–71 one additional study was by telephone and mail.105 Three studies were conducted in community settings—one in a nonclinical site,118,119 one in churches,107 and one on community streets.117 For one study, the place of service was not reported.16,127 The majority of studies did not report the educational materials used; one of these studies provided condoms as part of the intervention.117 Only four studies provided some description of the educational materials used during the intervention.64–66,117–119
Community health worker-participant interactions for injury prevention interventions. We included three studies in injury prevention (Table 6).101,102,126 Two took place primarily in the home101,102 and one on farms.126 Two studies involved the distribution of materials to improve safety;102,126 one study did not report the educational materials used.101 Two studies were of moderate intensity;101,102 one was of low intensity.126
Community health worker-participant interactions for maternal and child health interventions. Overall we included 15 studies in maternal and child health (Table 7). All of the studies occurred primarily in the home, but 1 had opportunities for interactions in health care clinics.77 Only 4 studies provided some description of educational materials used during the intervention;75–77,80–82,84 the remaining 11 did not report any details.67,68,71–74,78,79,83,85–87,128 All the maternal and child health studies were of high intensity.
Community health worker-participant interaction for cancer screening interventions. Overall 15 studies concerned cancer screening: 7 took place primarily in the home (visits or telephone )15,17–22,61–63,103,104,106 and 8 in community locations59,60,107–113,116,125 (Table 8).
Nine studies described some of the materials used during the intervention;15,17,18,59–63,103,109–113 six did not report the educational materials used.19–22,104,106–108,116,125 We found two studies of high intensity,17,18,61,62 seven of moderate intensity.15,63,106,109–112,116,125 and six studies of low intensity.19–22,59,60,103,104,107,108,113
Community health worker–participant interactions for chronic disease management interventions. Overall, 13 studies focused on chronic disease management (Table 9). Seven took place primarily in the home,23,88–100 2 in health care settings123,124 and 4 in community locations.27,114,120–122 Eight described some of the materials used during the intervention;27,93–98,100,114,123 five did not report the educational materials used.23,88–92,99,120–122,124 Two studies provided materials to households to reduce exposure to asthma triggers (bedding, vacuum cleaners, etc.). 96,97,100 Eight were of high intensity 27,88–98,100,120,121 and five studies were of moderate intensity.23,99,114,122–124
KQ 2: Outcomes of Community Health Worker Interventions
KQ 2 asks about the impact of CHWs on outcomes, particularly knowledge, behavior, satisfaction, health outcomes, and health care utilization. As noted earlier, the effect of CHW interventions will vary by clinical context (e.g., diagnosis or health concern), so as with KQ 1, we present results by clinical context for each of the outcomes described above. The areas of clinical concern are health promotion and disease prevention, injury prevention, maternal and child health, cancer screening, and chronic disease management.
We also assessed each study for quality; in general, we present results for higher quality studies first, followed by findings for moderate and then lower quality studies. We also give the level of intensity of the interaction between CHWs and participants and the type of comparator for each study, using the three intensity categories introduced in KQ 1. As noted there, the intensity of the interaction between CHWs and participants varied by clinical context. For example, maternal and child health interventions were solely high intensity whereas cancer screening studies ranged across high, medium, and low intensity. Because of this variation for cancer screening, we discuss those studies categorized first by intensity, then by quality. For all other clinical contexts, we did not find meaningful patterns by intensity of intervention, either because of lack of variation in intensity, or because the number of studies was insufficient to draw conclusions.
Variation in aims and clinical contexts of the studies, populations and settings, measures of health outcomes, and health care utilization information precluded quantitative synthesis of the results of studies. As with other questions, the number of articles exceeds the number of distinct studies. In all cases, tables list studies by quality (good, fair, then poor) and then alphabetically by last name of the first author of the article(s).
Outcomes for Health Promotion and Disease Prevention
Health promotion and disease prevention: pediatric immunizations. Study characteristics. Two RCTs, one good69,70 and one fair quality,68 and one poor-quality prospective cohort study (REACH-Futures71,72) examined outcomes of CHW interventions to improve pediatric immunization rates in inner cities (Table 10). The RCTs used moderate-intensity interventions and the cohort study used a high-intensity intervention.
Both RCTs used CHWs to provide reminder telephone calls for upcoming clinic appointments. The good-quality RCT, targeting children < 12 months of age in a county public health clinic in metropolitan Atlanta, had CHWs make home visits only if a child remained behind on his or her immunization schedule.69,70 Additionally, this study compared four groups of children receiving: (1) automated telephone call reminders, (2) CHW outreach, (3) a combination of a CHW and automated telephone call reminders, and (4) a control group defined by normal clinic procedure.69,70 Outcomes were assessed after 22 months.69,70
The fair-quality trial, targeting low-income children in Manhattan, also used CHWs to provide basic immunization education and referral, in addition to assisting in obtaining immunization services through a combination of telephone and home visits.68 It compared outcomes after 6 months for children receiving the CHW intervention with those for a control group comprising parents who were informed of their child’s immunization status at enrollment and instructed to reschedule the missed appointment.68
REACH-Futures, a prospective cohort study, compared a group receiving a high-intensity intervention of CHW and nurse visits with historic controls of nurse-only home visits.71,72 Monthly home visits started prenatally and ended at 1 year.71,72 We rated this study poor because of high potential for secular trends, given the time difference between the two groups, and for other confounding problems.71,72
Overview of results. These three studies68–72 evaluated the impact of CHWs on vaccine series completion rates and showed different CHW effectiveness. The good-quality study found no difference between groups receiving the CHW intervention and the control group.69,70 In contrast, the fair-quality study demonstrated that children in the CHW group were more up-to-date and less likely to be late for their immunizations than the controls.68 The control group for this study received more intervention directed at improving immunization rates, which would diminish the apparent effectiveness of the CHW. This study was more intensive than either of the other two projects (regular home visits or telephone calls over 6 months to ensure that requisite vaccines were received); this factor may have produced the difference in effectiveness between studies. REACH-Futures71,72 also found that the CHW-intervention group had a higher proportion of fully immunized participants at 12 months than did the historic controls who had received a nurse-only home visit.
Knowledge. No study reported outcomes for improved knowledge of pediatric immunization.
Behavior. No study reported outcomes for behavior changes.
Satisfaction. No study reported outcomes for satisfaction.
Health outcomes. All three studies evaluated immunization rates. The good-quality trial evaluated vaccine series completion rate from an immunization registry and found no difference between the CHW and control groups.69,70 The fair-quality trial found that children in the CHW arm were more up-to-date on immunizations than in the control arm (75 percent versus 54 percent, P = 0.03) and that fewer children were late for immunizations (18 percent versus 38 percent, P < 0.5).68 The poor-quality study evaluated vaccine series completion rates at 12 months and found that a higher proportion of children receiving the CHW and nurse home visits were up-to-date than historical controls (P < 0.001).71,72
Health care utilization. No study reported outcomes for health care utilization.
Health promotion and disease prevention: health promotion – Latina health. Study characteristics. Two RCTs, one fair66 and one poor quality,64,65 examined outcomes of CHW interventions in comparison with mailings for health promotion in Latinas (Table 11). The fair-quality study used a moderate-intensity CHW intervention in uninsured Hispanic women age 40 years and older living at the US-Mexico border (Agua Prieta, Sonora, Mexico, and Douglas, Arizona, United States) with the aim of increasing return to clinic for an annual preventive examination.66 It compared a group receiving CHW home visits in addition to reminder postcards with a group getting reminder postcards alone.
The poor-quality study, Secretos de la Buena Vida, used a high-intensity CHW model in the same target population living in San Diego County, California.64,65 It evaluated the effectiveness of weekly CHW home visits and telephone calls in addition to tailored print materials against that of tailored materials alone or off-the-shelf materials for changing dietary behavior. We rated this a poor-quality study because of a high potential for selection bias, measurement bias, and confounding.64,65
Overview of results. The fair study found that a moderate-intensity CHW intervention was more effective than a reminder postcard in increasing preventive exam appointments.66 The poor-quality study demonstrated that a high-intensity CHW intervention group was different from those receiving weekly tailored dietary printed material in terms of dietary intake immediately post-intervention. This difference was no longer apparent after 6 months, although all three groups improved.64,65
Knowledge. Neither study reported outcomes for improved knowledge of health promotion.
Behavior. The Secretos de la Buena Vida project examined behavioral changes.64,65 The CHW arm and the tailored print arm did not differ significantly at 6 and 12 months postintervention in dietary intake of fat or fiber, based on a validated measure for 24-hour diet recall.
Satisfaction. Neither study reported outcomes for satisfaction.
Health outcomes. Neither study reported outcomes for improved health.
Health care utilization. The fair-quality, moderate-intensity CHW study reported on the percentage of women returning to clinic for a second annual preventive examination.66 The CHW arm had a higher percentage of women returning than the postcard-only arm (65 percent versus 48 percent; RR, 1.35; 95percent CI, 0.95–1.92), but the difference was not statistically significant.
Health promotion and disease prevention: disease prevention. Study characteristics. Six studies, five RCTs16,67,105,107,118,119,127 and one prospective cohort study,117 examined outcomes of CHW interventions for disease prevention in underserved populations throughout the United States (Table 12). Two studies were both high intensity and fair quality;16,67,127 two studies were moderate intensity, one fair105 and one poor quality;118,119 and two studies were low intensity, one fair117 and one poor quality.107 Studies focused on a broad range of disorders, including cardiovascular disease prevention,105,118,119 diabetes prevention,16,127 HIV prevention,117 colorectal cancer prevention,107 and second-hand smoke exposure.67 Of the five RCTs, three were of fair quality16,67,105,127 and two were poor.107,118,119
The Missouri study was a fair-quality RCT evaluating a high-intensity CHW intervention focused on diabetes prevention in a low-income, African-American female population.16,127 This study compared 3 months of weekly sessions, alternating between group and individual sessions, targeting stages of change to tailor dietary patterns, with a control group that received a book to read.16,127 The San Diego study was a fair-quality RCT evaluating a high-intensity CHW intervention focused on decreasing secondary tobacco smoke exposure in Latino neighborhoods in San Diego County, California.67 The intervention consisted of six home and/or telephone visits by CHWs over 4 months using culturally tailored behavioral problem-solving techniques to reduce secondary tobacco smoke exposure; controls received no intervention.67 The Seattle, Washington, study was a fair-quality RCT evaluating moderate-intensity CHW assistance with medical followup against verbal advice to see a medical provider in low-income neighborhood participants who were found to have elevated blood pressure.105
The sole prospective cohort study, rated fair quality, evaluated the effectiveness of a low-intensity CHW intervention in HIV prevention by street outreach to at-risk community members in Louisiana compared with a control group in a neighborhood receiving no intervention.117
The poor-quality Baltimore, Maryland, trial evaluated a moderate-intensity intervention consisting of a nurse practitioner and CHW team at a nonclinical site with exercise equipment; CHWs provided dietary, smoking cessation, and exercise counseling.118,119 This strategy was compared with “enhanced” primary care, in which the same risk-specific materials and information on local programs were given to the intervention group and results and recommendations were provided to the patients’ primary care physicians. We rated it poor because of a high potential for measurement bias.118,119 The WATCH trial was a poor-quality RCT of low intensity conducted in rural, predominantly African-American churches in North Carolina.107 This study had four arms: (1) control churches offered a health education session and speakers not related to study objectives; (2) CHW intervention, consisting of organization and presentation of at least three church-wide activities on educating and enhancing support for healthy lifestyle and colorectal cancer screening; (3) four personalized computer-tailored newsletters and four targeted videotapes focused on healthy lifestyle and colorectal screening mailed bimonthly to participants’ homes; and (4) both the CHW and the videotape components.107
Overview of results. These six disease prevention studies reported on outcomes of knowledge, behavior, health outcomes, and health care utilization. Overall, four studies found that a CHW intervention was more effective in achieving outcomes than the respective control group.16,105,117–119,127 Two fair-quality studies (the Missouri trial16,127 and the prospective cohort study117) reported improved knowledge of the respective diseases in the CHW intervention as compared to respective controls. Two fair-quality studies (the Missouri trial16,127 and the prospective cohort study117) and one poor-quality study (the Baltimore trial118,119) demonstrated that moderate- and low-intensity CHW interventions were more effective than controls in changing health behaviors.
The two studies that targeted tobacco cessation found opposing results regarding CHW effectiveness.67,118,119 The fair-quality study (San Diego trial67) found no difference in smoking cessation between a high-intensity CHW intervention group and a group receiving nothing based on validated radioimmunoassay (RIA) of children’s hair for nicotine and cotinine. The poor-quality study (Baltimore trial118,119) found a significant difference between a moderate-intensity CHW intervention and enhanced usual care; however, this outcome was based on self-report. The fair-quality Seattle trial measured health care utilization and demonstrated that a moderate-intensity CHW intervention increased medical followup compared with only verbal advice to seek medical care for elevated blood pressure.105
Overall, most (four of the six) disease prevention studies demonstrated that various levels of CHW intervention intensity (low, moderate, or high) were more effective than the comparator, which ranged from nothing to enhanced usual clinical care, in changing a variety of outcomes.
Knowledge. Two fair-quality studies16,117,127 reported outcomes for improved knowledge of the respective diseases. The Missouri study16,127 found that participants in the high-intensity, diabetes-oriented CHW intervention, compared with a control group receiving a book to read, had an improved knowledge of label reading as assessed by an unvalidated questionnaire (P < 0.0001); this improvement remained statistically significant at 6-month followup. The prospective cohort study117 demonstrated that a low-intensity CHW street outreach program was effective at increasing knowledge of where to obtain free condoms as determined by an unvalidated questionnaire (90 percent versus 74 percent, odds ratio [OR], 3.2, P = 0.001).
Behavior. Five RCTs, three fair16,67,105,127 and two poor quality,107,118,119 examined a variety of behavioral changes. Three demonstrated CHW effectiveness16,105,118,119,127 and two67,107 showed no difference compared with their respective controls. The Missouri trial on diabetes prevention evaluated dietary change following high-intensity, CHW-led group and individual sessions;16,127 it found a reduction in fat intake with a validated food frequency questionnaire compared with intake in a control group (P < 0.0001). The San Diego trial, a high-intensity CHW intervention of home and telephone visits to reduce second-hand tobacco smoke exposure to children, found no difference from baseline by self-report or validated RIA of children’s hair for nicotine and cotinine.67 In contrast, the Baltimore trial evaluated a CHW intervention and found a difference in self-reported smoking cessation as compared to a standard of care group (16.2 percent reduction versus 7.0 percent, P < 0.001).118,119 Both groups reported less smoking, confirmed by measures of hair cotinine. The North Carolina trial did not show a difference in either fruit and vegetable intake or increased physical activity between intervention and control groups.107 The prospective cohort low-intensity study targeting HIV prevention demonstrated an increase in condom use reported in the intervention group (OR, 1.37; 95 percent CI, 1.20–1.56).117
Satisfaction. No study for health promotion evaluated outcomes focused on satisfaction.
Health outcomes. The Missouri trial found no difference within or between arms when comparing the high-intensity CHW intervention and the control group in terms of body weight and body mass index (BMI) at baseline (BMI 35.7 versus 35.3) and after 6 months (BMI 35.7 versus 35.4).16,127
Health care utilization. The Seattle trial evaluated self-reported medical provider followup within 90 days of determined elevated blood pressure.105 It demonstrated a higher rate of completed medical followup in the CHW group than in the control group (65.1 percent versus 46.7 percent, P = 0.001). The number needed to treat in order to bring 1 person to medical care was 5 (95 CI, 3–13).105
Outcomes for Injury Prevention
Injury prevention: home safety. Study characteristics. One fair-quality RCT101 and one poor-quality RCT randomized at the community level (called the Safe Block Project)102 assessed the effect of low-intensity CHW interventions on injury prevention in homes, either for children101 or for all ages.102 Both studies involved CHW home visits. The fair-quality RCT consisted of assessment of safety hazards and recommendations for appropriate products and practices compared with safety counseling in a pediatric clinic.101 The poor-quality RCT also included direct implementation of several safety features into homes compared with no intervention in control households; we rated this trial poor because of its high potential for measurement bias and not masking those who assessed outcomes.102
Overview of results. The fair-quality RCT showed no benefit to CHW intervention,101 but the poor-quality trial had mixed results102 (Table 13.). Significant benefit was seen for household features that did not require participants to change behaviors (e.g., continued presence of a smoke detector, as installed in intervention homes); conversely, no benefit was observed for other household features that did require behavior change (e.g., maintaining a working light bulb in stairways).
Knowledge. Neither study assessed knowledge-related outcomes.
Behavior. In the fair-quality RCT,101 groups did not differ significantly in maintaining adequate stairway lighting (83.1 percent versus 80.1 percent; adjusted odds ratio [AOR], 0.90; 95 percent CI, 0.69–1.16) or in following any of the home safety practices assessed. Hot water temperature control and presence of ipecac increased from baseline in both groups, but presence of a working smoke alarm, use of safety gates on stairs, and latching or locking of poisons declined from baseline. In the poor-quality trial,102 following the CHW intervention a significantly higher proportion of households continued to have ipecac (which was recommended at the time of the study for households with young children) (81.0 percent versus 9.8 percent; AOR, 0.04; 95 percent CI, 0.02 to 0.07) and smoke detectors (96.0 percent versus 77 percent; AOR, 0.14; 95 percent CI, 0.09 to 0.20) than did controls. These interventions were provided by the CHWs and required no behavior change by participants. In contrast, intervention households were actually less likely than control households to have retained hot water temperature controls (63.2 percent versus 73.2 percent; AOR, 1.73; 95 percent CI, 1.39 to 2.15).
Satisfaction. Neither study assessed satisfaction.
Health outcomes. Neither study assessed direct health outcomes.
Health care utilization. Neither study assessed health care utilization.
Injury prevention: workplace safety. Study characteristics . One prospective cohort study, rated poor quality for high potential for selection and measurement bias and lack of description of baseline characteristics, examined the effect of a low-intensity CHW intervention for migrant farm workers to prevent work-related eye injury. The CHW intervention involved distribution of protective eyewear either with or without specific training provided by the CHWs; it was compared to distribution of eye protection not involving CHWs. Outcomes were assessed during the same growing season in parts of the Midwest.
Knowledge. Knowledge was not assessed.
Behavior. The presence of any CHW component related to receiving protective eyewear was significantly associated with increased self-report of continued use of the eyewear on a 5-point Likert scale compared with having received the eyewear without CHW involvement (difference in average change in Likert scale value 0.6452, P < 0.01). Incorporation of CHW-led training was associated with greater self-reported eyewear use compared with CHW eyewear distribution alone (difference in average change 0.7663, P < 0.01) and with no CHW involvement (difference in average change 0.5241, P = 0.03). Observed use of eyewear increased in all groups during the study period (CHW trained 1.1 to 36 percent; CHW distributed 0 to 5.2 percent; no CHW 0 to 14 percent, P-value not reported).
Satisfaction. Satisfaction was not assessed.
Health outcomes. Although the investigators measured the incidence of pterygium, they did not compare groups on this variable and in fact reported it as only inadequately identified.
Health care utilization. No measure of health care utilization was reported.
Outcomes for Maternal and Child Health
Maternal and child health: overview. We identified 15 studies that met inclusion criteria and involved maternal or child health outcomes (or both). All the studies utilized high-intensity interventions, usually involving some series of home visits. All but 1 study were rated either fair (8 studies) or poor (6 studies). The 1 good-quality study found no significant differences associated with interventions employing CHWs. Among the other studies, results were mixed, some showing benefit of CHW interventions and some showing no effect attributable to CHWs. This distribution was found in both fair- and poor-quality studies. Significant associations were most commonly found for existing conditions (e.g., phenylketonuria [PKU] or failure to thrive) rather than primary prevention and in the area of health care utilization (e.g., immunization rates) and behavior (e.g., parenting measures).
Maternal and child health: prenatal care and perinatal outcomes. Study characteristics. Six studies assessed prenatal care and perinatal outcomes associated with CHWs.71,72,77,79,83,86,87 Of these, three were rated fair quality: one RCT involving prenatal care in Cleveland79 and two cohort studies (one on the Resource Mothers Program for Maternal PKU83 and one evaluating REACH-Futures72). The remaining three studies, rated poor, included one RCT on promotion of breastfeeding in African-American mothers in Baltimore,77 rated poor for high attrition and lack of specific or validated outcome measures; one cohort study (the Baby Love Maternal Outreach Worker study31,32), rated poor for high attrition, high potential for selection bias and confounding, and lack of specific or validated outcome measures; and a second study on REACH-Futures71 rated poor for high potential for secular trend and other confounding.
Most studies focused on interventions for low-income families, usually from racial or ethnic minority groups. Most CHW interventions involved home visits. The Resource Mothers Program for Maternal PKU83 involved coaching in activities of daily living unique to mothers with PKU infants including meal planning and medical recommendations concerning pregnancy. The Maternal Outreach Worker program also provided direct assistance to families for obtaining benefits and services.86,87
Studies generally compared outcomes for families receiving CHW interventions with outcomes for those receiving usual clinical care (Table 15). The Baltimore breastfeeding study compared CHW intervention with video and other literature and against both interventions combined;77 the Resource Mothers Program83 used as controls mothers who had completed pregnancy in the 5 years before the start of the program; and REACH-Futures71,72 used historic controls of nurse home visits. Outcomes were typically assessed months to years after the interventions.
Overview of results. Improvements over usual care were demonstrated to be associated with CHWs in breastfeeding,77 maternal control of PKU,83 and prenatal care.86,87 However, birth outcomes in mothers with PKU,83 low birth weight incidence,86,87 continuation of breastfeeding,77 and overall presence of infant health problems71 were not significantly improved by use of CHWs compared with usual care77,83,86,87 or with health professional intervention.71
Knowledge. No study measured knowledge-related outcomes.
Behavior. No study assessed behavior change.
Health outcomes. Peer CHW counseling in the Baltimore study was associated with greater initiation of breastfeeding than standard care (OR, 3.84; 95 percent CI, 1.44–10.21), but the statistically significant difference between groups in the proportion of participants still breastfeeding by 7 to 10 days disappeared.77 For the Resource Mothers Program,83 mothers receiving the CHW intervention needed less time to reach metabolic control (blood phenylalanine level consistently below 10 mg/dL) than those who had not received the intervention (8.5 weeks versus 16 weeks, P < 0.05). The head circumference of infants born to participating mothers did not differ significantly between cohorts (mean Z-score of head circumference: intervention −0.56; 95 percent CI, −0.88 - −0.24 versus control –1.4; 95 percent CI, −1.56 - −1.2; P = 0.08). The Maternal Outreach Workers program demonstrated a trend toward lower incidence of adequate prenatal care for African-American women receiving CHW intervention than for controls (significance not reported);86,87 neither the observed nor the expected incidences of low birth weight or very low birth weight infants differed significantly. REACH-Futures found no difference between CHW intervention and controls in incidence of neonatal or postneonatal infant health problems.71
Health care utilization. The Cleveland study showed a significant increase in the ratio of actual to expected numbers of prenatal visits for women receiving CHW intervention (P = 0.029);79 the investigators did not compare the intervention findings to those from women in the control group.
Other. The fair-quality analysis from REACH-Futures found that CHW home visits were more likely than nurse home visits to include identification of problems in women’s health (P = 0.01), deficits in well-child care (P = 0.02), parenting issues (P = 0.02), and socioeconomic issues (P < 0.01) and that participants were more likely to receive problem-solving services (P < 0.01).72 However, CHWs were less likely than nurses to provide emotional support services (P < 0.01) or to place referrals for women’s health (P = 0.01), well-woman care (P = 0.02), emotional/interpersonal support (P < 0.01), parental support (P < 0.01), or socioeconomic issues (P < 0.01).
Maternal and child health: Child development. Study characteristics. Four studies considered the impact of CHWs on child development (Table 16). Three were rated fair quality and one poor quality; all used high-intensity interventions. One RCT focused on children with nonorganic failure to thrive in Baltimore, Maryland;75,76 another RCT examined the Home Visitation 2000 program in Denver, Colorado;80–82 and a cohort study involved the Resource Mothers Program for Maternal PKU in New England.83 The RCT assessing the Hawaii Healthy Start Program78,128 was rated poor for high potential for site-specific bias.
Overview of results. Variation in timing and specific outcomes among studies precludes much summarization of results. Two of the studies demonstrated some significant benefit of CHW intervention over usual care; the other two showed no significant difference between CHW intervention and controls. The failure-to-thrive study demonstrated that the CHW home visiting program was effective in mitigating declines in cognitive and motor development, but not language, if implemented during the first year of life (Table 16).75,76 The PKU Resource Mothers Program study found higher mental development for infants born to mothers who participated than for those born to historic controls.83 By contrast, the Home Visitation 2000 trial showed more improvement in language development with nurse visits rather than CHWs,80–82 and the Hawaii trial found no difference in mental or psychomotor development between children receiving CHW intervention and controls.
Knowledge. No study assessed knowledge about child development issues.
Behavior. No study included health behaviors in the outcomes measured.
Satisfaction. No study considered satisfaction outcomes.
Health outcomes. All four studies examined various health outcomes. In the Baltimore, Maryland, failure-to-thrive study, the decline in cognitive development over 1 year as measured by the Bayley Scales of Infant Development was less severe for the home intervention group than for the clinic-only group (P = 0.02) for children recruited during infancy. Groups of children recruited at older ages did not differ using the Battelle Developmental Inventory, although all groups demonstrated some degree of decline in cognitive function. Whether this decline was attributable to failure to thrive or to some other factor was not assessed in the study. Children in the intervention group showed less severe decline in receptive and expressive language than did age-matched controls (P = 0.05), but all groups experienced relative declines in language over the course of the study. All groups showed significant improvements in weight for age, weight for height, and height for age, but the groups did not differ significantly.
The Home Visitation 2000 study in Denver, Colorado, found slightly greater improvement over controls with nurse home visits than with CHW visits for the Preschool Language Scales at 21 months and the Mental Development Index at 24 months.80–82
Infants in the intervention cohort of the Resource Mothers Program in New England had higher mean Bayley Developmental Quotient (mental scale) values than those in the control cohort (108 versus 95) at 12 months of age (P < 0.05).83
At 2 years postintervention, children in the Hawaii Healthy Start Program78,128 who received CHW intervention had a mean Bayley Mental Development Index score of 90.0 versus 89.2 for controls (P = 0.60) and a Psychomotor Development Index score of 92.1 versus 90.4 for controls (P = 0.12).
Health care utilization. No study assessed health care utilization.
Maternal and child health: Environment conducive to child well-being. Study characteristics. Factors contributing to an environment conducive to the health and well-being of children were assessed directly in 10 studies; 6 rated as fair quality and 4 as poor quality. The five fair-quality RCTs covered the following populations and interventions: smokers in San Diego;67 low-income urban children with nonorganic failure to thrive;75,76 the Parent to Parent Network for mothers of children with chronic conditions;85 a trial targeting children in New York with missed immunization visits;68 a trial involving drug-using mothers in Maryland;84 and the Home Visitation 2000 RCT.80–82 Finally, of the four poor-quality studies, two RCTs (both involving the Child-Parent Enrichment Project, or CPEP73,74) were rated poor for lack of relevant outcome measures; the Hawaii Healthy Start Program78,128 was rated poor for high potential for site-specific bias; and on the REACH-Futures trial71 was rated poor because of high potential for secular trend and for other confounding.
Overview of results. The variety of outcomes assessed by the studies precludes much summary of results. Of the 10 studies in this category, only 4 reported significantly beneficial outcomes for CHWs over usual care.
The New York study68 and REACH-Futures trial71 did find CHW-associated improvements in immunization status. Home Visitation 2000 showed greater improvement with nurse than with CHW interventions for mother-infant interaction, home environment, and tobacco smoke exposure.80–82 The Hawaii study found that CHW intervention significantly increased appropriate parental coping and discipline methods and decreased injuries from partner-related violence.78,128
As to the remaining studies: the San Diego study found no significant impact by CHWs on exposure to environmental tobacco smoke among children of smokers.67 The failure-to-thrive study found no effect of CHWs on outcomes related to home environment or parenting behavior.75,76 The Parent to Parent Network study showed no significant difference between intervention and control groups for maternal psychiatric well-being postintervention;85 however, the results were potentially confounded by differences at baseline. No differences were found in the Maryland study84 for maternal drug use or mother-child interaction. Other studies on substance abuse, child maltreatment, and improving psychiatric outcomes among caregivers of children with chronic diseases also did not report significant differences between study arms.74,84
Knowledge. No study assessed measures of knowledge.
Behavior. The failure-to-thrive study found no differences between groups for parent-child interaction behavior during feeding using a modified Parent Child Early Relational Assessment.75,76 It did show improved interactive communication with parents during feeding among children over time for all groups (P < 0.001), but no differences were apparent according to intervention status. Developmental appropriateness of the home environment, as assessed postintervention by the Home Observation for Measurement of the Environment Scales, was slightly higher for the CHW intervention group than for the clinic-only group (31.6 [SD 3.6] versus 29.3 [SD 4.2] for infants; 32.4 [SD 5.1] versus 30.3 [SD 5.7] for older children; P = 0.05 [significance not reported by age strata]). However, no baseline scores were reported for this measure to ascertain the true effect of CHWs.
In the Maryland study on substance-abusing mothers, self-reported postintervention substance use was similar for mothers receiving CHW interventions and for those in the control group (65 percent versus 68 percent for alcohol, 46 percent versus 44 percent for cocaine and/or heroin, and 25 percent versus 38 percent for marijuana; P ≥ 0.1).84
The Hawaii study found that parents who received CHW intervention had a greater postintervention use of nonviolent discipline strategies (see Table 16), reported less parenting-related stress, and had higher parenting efficacy scores than those receiving usual care alone.78,128
Satisfaction. No study assessed satisfaction outcomes.
Health outcomes. Among children of smokers in the San Diego study,67 no reduction was seen in parental report of children’s tobacco exposure or in nicotine or cotinine levels in children’s hair for either CHW or control participants. The Parent to Parent Network demonstrated no difference between groups in postintervention Psychiatric Symptom Index scores (intervention 22.1 versus control 20.1).85 However, the baseline score for the intervention group was significantly higher than for the control group (24.1 versus 20.3, respectively; P < 0.05). Adjustment for this baseline difference revealed a greater degree of improvement in the intervention group than in controls, except for the depression subscale, which was improved in both groups. However, whether this reflected true improvement attributable to CHWs or was simply a regression to the mean could not be determined. The New York study showed that children receiving CHW intervention were more likely than control children to be current on their immunizations (P = 0.03) and less likely to have received immunizations behind schedule (P < 0.05) (Table 17).68
The Maryland study found infant warmth (on a 5-point scale) to be equal for those receiving CHW interventions and controls (2.5, SD 0.4 for both groups).84
Home Environment 2000 demonstrated more improvement over controls in mother-infant interaction and in home environment for nurse home visits (least squares mean 1.32, P ≤ 0.05) than for CHW visits (least squares mean 1.16, P < 0.1).80–82 Among participating families with mothers who smoked, maternal urine cotinine was reduced in all groups; those receiving nurse home visits had a significantly greater degree of reduction than those receiving CHW visits (nurse versus control −246.68 ng/dL; 95 percent CI, −466.19 to −27.16); CHW versus control −76.19 ng/dL; 95 percent CI, −302.21 to −149.82; P ≤ 0.05).
The studies from Contra Costa, California, found no significant difference between groups on the Child Abuse Potential Inventory postintervention (Table 17); both groups showed improvement and no difference in reported cases of child maltreatment.73,74
The Hawaii study78,128 demonstrated no difference between groups for maternal life skills (Table 17), maternal social support, maternal substance use, maternal depressive symptoms, or incidence of poor general mental health among mothers at 2 years postintervention. Neither home learning environment nor parent-child interactions differed between groups at 2 years. The investigators did not report how each of these measures compared with baseline values. The study did show lower incidence of injuries attributable to partner-related violence among families receiving CHW intervention (P = 0.03), but no differences in reported or confirmed cases of child maltreatment.
Health care utilization. Children receiving CHW intervention in the Hawaii study were no more likely than those receiving usual care to have a primary care provider (P = 0.09) (Table 17), to have received the recommended number of well-child visits (P = 0.95), or to be current on immunization status (P = 0.45).78,128
Outcomes for Cancer Screening
Cancer screening. Study Characteristics. A total of 15 studies (24 citations) examined outcomes of CHW interventions for improving breast, cervical, or colorectal cancer screening.15,17–22,59–63,103,104,106–113,116,125 Information on these studies is spread across multiple tables, depending on the specific focus: improving knowledge, changing behavior, breast self-examination, Pap smears, mammography, clinical breast examination, and colorectal cancer screening.
Of these studies, 10 are RCTs15,17–22,61–63,103,104,106–113 and 5 are observational studies.59,60,108,113,116,125 The RCTs include three randomized by communities103 or churches.19–22,104,107 Of the five observational studies, one was a quasi-experimental controlled cohort,125 two were prospective cohorts,108,113 one used retrospective records,59,60 and one used repeated cross-sectional survey of women attending beauty salons randomly assigned to experimental and control groups.116 The studies spanned the quality range as well: two were of good quality,17,18,103 seven of fair quality,59–63,106,108,109,125 and six of poor quality.15,19–22,104,107,110–113,116
As noted in our section on KQ 1, seven studies used low-intensity CHW models, 6 used moderate-intensity interventions, and two used high-intensity interventions. Six studies included more than two arms. Studies compared the CHW arm with a variety of alternatives, including no intervention or usual care (6 studies), mail (3 studies), community interventions (4 studies), CHWs in a lesser capacity (2 studies), and CHWs in combination with other interventions (2 studies).
Studies used varied definitions of outcomes. The greatest commonality was reporting on utilization of cancer screening tests such as mammography, clinical breast examination, Pap smears, and colorectal cancer screening. Of the 15 studies, 13 reported on changes in rates of utilization, but they varied in their specific definitions (ever use, use in the past 3 months, 1 year, 2 years, and so on).15,17–22,59–63,104,106–113,125 With the exception of 3 studies examining Medicaid or medical records for mammography use,17,18,59,60,63 all relied solely on self-report.
Overview of results. Together, the 15 studies suggest limited evidence of improvement in knowledge in the CHW arm compared with alternative approaches; they present conflicting findings on the effect of CHWs on planned and actual health behaviors, specifically breast self-examination. The volume of evidence on these outcomes is limited; the quality and design of the studies limits the interpretation of available evidence.
Unlike most of the other subsections dealing with other purposes for CHW strategies, cancer screening studies used high-, moderate-, and low-intensity interventions. Enough studies and evidence are available to permit some analysis by the intensity variable as it relates to Pap smears and mammography. Summary tables for these two outcomes are therefore presented by intensity (low, then moderate, then high), followed by quality, and then alphabetical order, by last name of first author(s); for all other sections, we present studies by quality, and then alphabetical order, by last name of first author(s).
Regarding health care utilization, our findings from this limited evidence do not support the conclusion that CHW interventions are more effective in comparison with other alternatives in raising the rates of clinical breast examination or colorectal cancer screening. More substantial evidence exists on Pap smears and mammography. The CHW approach is at least as effective as the alternative in improving Pap smear rates, but it is more effective than the alternative only in limited circumstances of low- and moderate-intensity interventions. With respect to mammography rates, studies demonstrated significantly greater improvements in the CHW arm compared with the alternative (no intervention, mail, print, or minimal CHW) in either the entire sample or in subsamples.
Knowledge. Two studies (three articles; Table 18) examined changes in knowledge and found limited evidence of improvement for the CHW arm.17,18,109 A good-quality, high-intensity study in North Carolina measured knowledge for 12 individual measures on breast cancer and a composite score.17,18 The studies together suggest improvements in the CHW arm, although the results are not consistent on the relative benefit of the CHW arm versus the alternative. Although differences between the CHW and the comparison arm (mail intervention) were not statistically significant for the composite measure of knowledge, the study reported significant different improvements favoring the CHW arm on two individual items measuring knowledge. Both arms demonstrated improvements in other measures, but these improvements were not statistically significantly different. A second study, of fair quality and moderate intensity in California, found significantly different improvements on two measures of knowledge, favoring the CHW arm compared with the media intervention arm.109
Behavior: planned testing. Two studies, one of fair quality and low intensity63 and the other of poor quality and moderate intensity,116 provide contradictory findings on the effect of CHWs on planned behavior Table 19). The fair-quality study compared a CHW arm with direct and usual care; differences in the rate of planned Pap smear tests favoring the CHW arm were statistically significant compared with either direct mail or usual care.63 The poor-quality study reported no differences among study arms. However, the design of the study, which involved repeated cross-sections in salons randomly assigned to experimental and control status in which experimental salons offered barrier-specific counseling, was not measuring changes in intent over time; rather, it was concerned with differences in a cross-sectional sample. Low penetration combined with contamination across the samples (as suggested by the 37 percent and 10 percent of the sample reporting breast health messages at control sites and experimental sites, respectively) could have diluted the effects of the intervention.116
Behavior: breast self-examination. Five studies (eight citations; Table 20) reported on changes in self-breast examination as outcomes of CHW interventions.61,62,108,110–112,116,125 Of these five studies, three were of fair quality61,62,108,125 and two of poor quality.110–112,116 They included one high-intensity,61,62 three moderate-intensity,110–112,116,125 and one low-intensity study.108
These studies provide conflicting evidence of the effectiveness of the CHW approach, either in comparison with an alternative or over time independent of a comparison. Two studies reported significant differences between the CHW arm and an alternative (low-intensity CHW, mailed intervention, delayed intervention, or no intervention).108,110–112 The same two studies also provided evidence of significant differences between baseline and followup for the CHW arm.108,110–112 A third study employed repeated cross-sectional measurements and reported higher rates in the followup assessment but these were not statistically significant.116 The fourth study failed to find any improvements over time.61,62 The fifth study found reported conflicting results for the two selected measures.
Of the three fair-quality studies, the high-intensity study compared the CHW arm with a mailed intervention,61,62 the moderate-intensity study compared the CHW arm (outreach) to no-intervention arm, and the low-intensity CHW arm compared the CHW arm to a delayed intervention.108 The high-intensity study found no significant improvements over time in either arm, or between arms, except when the sample was restricted to a much reduced subsample who were available at followup and were not on the recommended screening schedule.61,62 The moderate-intensity study found improvements in the intervention arm over time for ever use of breast self-examination, but no significant differences in the control arm, but also found opposite effects for another measure: monthly breast self-examinations, with significant differences in the control arm over time, but not the intervention arm The low-intensity study found that the CHW arm resulted in significant improvements over time compared with the delayed-intervention arm.108 However, baseline differences between the two arms were large; significant differences between the two arms could have resulted from ceiling effects.
Of the two poor-quality studies, one moderate-intensity intervention compared a more intense CHW arm with a less intense CHW arm;110–112 the other moderate-intensity intervention compared the CHW arm with a no-intervention control.116 In the former study, the two arms differed significantly through 1-year followup but not at the 2-year followup.110–112 In the latter study, rates of breast self-examination were higher in followup interviews than in baseline interviews, but the differences between the arms was not statistically significant.116
Satisfaction. No study reported outcomes for satisfaction.
Health outcomes: No study reported on health outcomes.
Health care utilization: Pap smears. The evidence on the effectiveness of CHW interventions draws upon six studies (nine articles; Table 21).17,18,61–63,110–112 Most studies demonstrate that the CHW arm is as effective as the alternative in improving Pap smear rates. CHWs were notmore effective than mailed interventions in high-intensity interventions. They were more effective than the alternative in limited circumstances involving low- or moderate-intensity intervention in three of four studies. Because intensity may, thus, actually be an important policy variable for analyzing use of Pap smears, we present information on Pap smear use ordered first by intensity and then by the quality of the studies.
One low-intensity trial, of fair quality, compared CHWs with direct mail and with usual care in cities in Washington and British Columbia.63 CHWs were more effective than either alternative in increasing Pap smear rates, using both self-report and medical records.
Of the three moderate-intensity interventions, one in Santa Clara County, California, was of fair quality109, a second in San Francisco and Contra Costa was also fair quality,125 and the third in San Diego County, California, was rated poor quality.110–112 The fair-quality study in Santa Clara compared CHWs with a media intervention; CHWs were effective in increasing rates of Pap smears.109 The fair-quality study in San Francisco and Contra Costa found no statistically significant difference in changes in self-reported Pap smears between residents of intervention and control communities.125 The poor-quality study compared a higher-intensity CHW arm focusing on cancer control with a lower-intensity CHW arm.110–112 Although both arms demonstrated effectiveness compared with baseline values, participation in the more intense arm did not affect use of Pap smears compared with the less intense arm in the short term, but it did demonstrate effectiveness in the longer term (at 1- and 2-year followups). Followups were marked by high dropout rates, however, so the effectiveness in the longer term could be explained by selection bias.
Two high-intensity trials, one good-quality study in North Carolina,17,18 and one fair-quality study among inner-city African Americans (location unspecified),61,62 compared CHWs to mailed interventions. These two studies reported consistent results failing to demonstrate effectiveness of CHWs in improving Pap smear use compared with mailed interventions, but both studies showed that both arms demonstrated improvement compared with baseline values.
Health care utilization: mammography. Eleven studies (21 articles; Table 22), provide evidence on the effectiveness of CHW intervention with respect to breast cancer screening by mammography.15,17–22,59–62,103,104,108,110–113,116 Eight of these studies demonstrated significantly greater improvements in the CHW arm compared with the alternative (no intervention, mail, print, or minimal CHW) in either the entire sample or in subsamples.17–22,59–62,103,104,108,110–113 Two of three studies reporting nonsignificant differences between the CHW arm and the alternative were moderate-intensity, poor-quality studies comparing CHWs with no intervention;15,116 one of these studies reported nonsignificant differences between the CHW arm and the control, favoring the CHW arm. The third was a moderate-intensity fair-quality study comparing the effect of CHW interventions with controls at the community level.125 As with use of Pap smears, intensity may be a relevant analytic variable, so we report findings below first by intensity, then by study quality.
Four studies did not report changes over time;15,17–22,103,104 one study failed to show improvement in the intervention area,125 and the remaining six studies all demonstrated some improvement in the control arm (no intervention, delayed intervention, mail, print, or minimal CHW), although the improvement was not statistically significant.
Studies conducting subgroup analyses demonstrated that CHW interventions can provide benefits for subpopulations. Four studies provide evidence that CHW interventions are likely to be more beneficial than alternative interventions in low-income, minority populations with some health care barriers.19–22,60,103,104,113
Low-intensity interventions generally compared CHW with minimal to no intervention, We identified five such studies, one good-quality,103 two fair,59,60,108 and two poor.19–22,104,113 Collectively CHWs were generally effective in raising mammography rates, but with potentially greater effects in subpopulations.
The good study from Washington State, comparing a no-intervention control group with CHW groups receiving community activities, individual counseling, or a combination of community activities and individual counseling found that all the CHW intervention arms had higher rates of new users than the no-intervention control, but the study did not find significantly greater effectiveness of CHW arms in comparison with a no-intervention control.103 The community activities arm appeared to be more effective than a no-intervention control in preventing relapse (that is, in ensuring that regular users or women who were adherent to recommended screening guidelines at baseline continued to be adherent at followup) than in enrolling new users.
One fair-quality study involved Latinas in Colorado enrolled in Medicare or Medicaid fee-for-service, or three other health maintenance organizations (Kaiser Permanente of Colorado, Access, and Anthem Blue Cross and Blue Shield).59 It found nonsignificant and modest differences in mammography screening rates in unadjusted analyses that compared a CHW intervention with a printed intervention. In adjusted analysis, the difference between the two arms was statistically significant, favoring the CHW arm. The other fair-quality study, in Arkansas, also reported significantly greater improvements in self-reported use of mammography in the CHW arm compared with a delayed intervention arm. However, the two groups differed significantly at baseline, with higher rates of ever-use of mammography reported in the control group; thus a ceiling effect limiting improvements in the control group cannot be ruled out.108
Four moderate-intensity interventions, one of fair quality125 and three of poor quality,15,110–112,116 reported outcomes for self-reported mammography use. The fair-quality study in San Francisco and Contra Costa found no statistically significant difference in changes in self-reported mammography between intervention communities and control communities.125 One study in New York, which compared CHW with no-intervention controls, found no significant differences between intervention and control arms after the intervention in use of mammography during the prior 3 months.116 Both studies described measure effects at the community level rather than at the individual level. Low penetration of the intervention and potential contamination between experimental and control samples limit the interpretation of the results.
Two other studies, both assessed as moderate intensity overall, compared higher-intensity CHW to lower-intensity CHW intervention;15,110–112 they both reported improvements in both arms. Only the study in San Diego County, California, found significant differences; it demonstrated that the relatively more intense arm was more effective in the 3-to 6-month period following the intervention.110–112 These improvements were not consistently significantly different between the two arms over the long run (1-and 2-year followups) for a reduced and potentially self-selected subsample.
Two high-intensity trials, one good-quality17,18 and one fair-quality,61,62 both compared CHWs to mailed interventions and reported improvements in the CHW arms of their studies. Only the good-quality study (using Medicaid records from North Carolina) found significant differences in mammography rates between the CHW arm and the mailed intervention arm.17,18 The fair-quality study, using self-reported mammography among inner-city African Americans (location unspecified), did not find any significant differences for the overall sample using intention-to-treat analysis, but it did report significant differences when analysis was limited to a potentially biased subsample of respondents available at followup.62,129
Four studies found evidence of effect modification in subgroup analysis.19–22,60,103,104,113 The evidence is derived from low-intensity studies of varying quality. The good-quality study from Washington found that CHW intervention arms were more effective than a control arm in subgroups: among regular users (women adherent at baseline), the CHW intervention arms showed significantly greater rates of mammography use among women who needed less than 2 hours to schedule a medical appointment.103 In the same study, subgroup analysis for under-users (women who were not adherent at baseline) found that the CHW interventions were significantly more effective than the no-intervention control among women without female doctors or insurance. These subgroup findings suggest that the CHW approach is effective in addressing some, but not all, access barriers to the use of mammography.
The fair-quality study from Colorado59,60 reported weak but slightly more powerful effects of the CHW approach compared with a printed intervention approach in increasing mammography rates among Medicaid-enrolled Latinas compared with non-Latina whites (P = 0.07 for Latinas, and P = 0.10 for non-Latina whites).60 Similarly, the poor-quality studies also suggested subgroup effects. One study found CHWs to be more effective than a no-intervention control group in increasing rates of self-reported mammography for the overall sample and in groups with incomes below $12,000, but not in groups with incomes equal to or exceeding $12,000.113 Another found that the CHW approach was more effective than with a minimal intervention approach in ensuring conversion to adherence among under-users rather than in maintaining adherence among regular or adherent users.19–22,104
Health care utilization: clinical breast examination. Four studies reporting on clinical breast examination (seven articles; Table 23)61,62,110–112,116 included a high-intensity and three moderate-intensity interventions. Two of these studies were of fair quality;61,62,125 the other two were rated poor.110–112,116 Together the studies suggest that CHW interventions are not effective in comparison with other alternatives, although two studies that provide information on changes between baseline and followup found that the CHW arm results in improvements over time.
The fair-quality high-intensity trial found no differences between the CHW arm and a mailed intervention, with the exception of a reduced and possibly selective sample of respondents only at followup.61,62 The fair-quality moderate-intensity study found no difference over time in most measures of self-reported clinical breast examination in intervention communities or control communities.125 Of the two poor-quality moderate-intensity studies, one trial compared a more intense CHW arm with a less intense CHW arm110–112 and the cross-sectional study compared it with a no-intervention arm.116 Neither study reported significant differences, although the women in the more intense CHW arm of the trial did report higher rates of clinical breast examination after the intervention.
Health care utilization: colorectal cancer screening. Two studies, one of moderate intensity and fair quality, and another of low intensity and poor quality compared three groups on outcomes for fecal occult blood tests (FOBT) and other colorectal cancer screening tests (Table 24).107 In the fair-quality moderate-intensity intervention, patients who received navigation services had higher rates of FOBT after three months of services than patients who received usual care, but these differences were not statistically significant. Patients receiving navigation services were significantly more likely than controls to have set an endoscopy appointment at three months and kept it by six months after the intervention.106 The low-intensity poor-quality study reported that rates of FOBT were higher in the CHW arm over time; however, the CHW arm and the comparison arms of a no-intervention control or of tailored print and videotapes did not differ significantly. The study reported no benefit of the intervention for other colorectal screening tests.
Outcomes for Chronic Disease Management
Chronic disease management: diabetes mellitus. Study characteristics. Four studies (eight articles; Table 25), three RCTs,27,88–92,124 and one prospective cohort study93 examined outcomes of CHW interventions for diabetes care among underserved minority populations with type 2 diabetes mellitus. All studies were rated fair quality. Three studies27,88–93 used a high-intensity intervention; one study124 used a moderate-intensity intervention.
The 6-month RCT conducted in Texas used a high-intensity intervention for Mexican Americans that compared eight weekly, 2-hour group classes with promotoras to usual care plus educational pamphlets.27 The RCT in New York City used a moderate-intensity intervention for inner-city Hispanics and African Americans that evaluated the use of CHWs as clinic liaisons compared with nurse-patient encounters.124 The Project Sugar trial RCT in Baltimore, Maryland, compared several high-intensity interventions in inner-city African Americans with type 2 diabetes: (1) CHW face-to-face home visits and telephone contact, (2) nurse care manager intervention, (3) a combined nurse care manager and CHW, and (4) standard clinical care with an additional quarterly diabetes newsletter.88–92 The prospective cohort study in Hawaii examined a high-intensity intervention comparing CHW diabetes case management, including home visits, in addition to a multidisciplinary team, with usual clinical care involving a multidisciplinary team approach.93 Heterogeneity of population, study designs, interventions, and outcomes preclude quantitative synthesis of results.
Overview of results. Of these four studies on diabetes management, two studies found the CHW intervention to be beneficial in decreasing hemoglobin A1c (HgbA1c) as compared with usual care;27,93 conversely, two studies found no difference between groups in mean change from baseline in HgbA1c.88–92,124 The Texas study also evaluated outcomes of knowledge and found that the CHW intervention was effective compared with usual clinical care in increasing diabetes knowledge.27 The Hawaii study found that diabetes case management by a CHW in conjunction with a multidisciplinary team was more effective at decreasing HgbA1c than a multidisciplinary team alone.93 The New York study demonstrated that a CHW liaison was more effective than usual clinical care in behavioral changes leading to program completion rates.124 Project Sugar, a high-intensity study, found significant changes from baseline within, but not between, groups for various health outcomes.88–92
Knowledge. The Texas study evaluated outcomes for improved knowledge at 6 months in diabetic patients following eight weekly CHW-led group classes in Mexican Americans.27 A validated tool, the bilingual Diabetes Knowledge Questionnaire (DKQ), showed a difference between arms, with an improved score in the CHW group compared with the group receiving usual care plus educational pamphlets (P < 0.002).27
Behavior. Project Sugar evaluated dietary risk scores (which identifies positive as well as problematic dietary behaviors and measures potential barriers to dietary change). Scores improved across all CHW arms as compared with the usual clinical care group following a high-intensity CHW intervention (all CHW arms versus usual clinical care [score ± standard deviation]: −2.4 ± 1.99 versus −3.45 ± 1.87 versus −2.13 ± 1.92; P not reported).88–92 The New York study demonstrated an increased proportion of completion of a diabetes education program after a low-intensity CHW intervention compared with usual clinical care (80 percent versus 47 percent, P = 0.01).124
Satisfaction. No study reported outcomes about satisfaction with diabetes care.
Health outcomes. The Texas trial demonstrated better improvement in diabetes control (measured by mean change in HgbA1c) in the high-intensity CHW intervention group than in the usual care group after 6 months (P < 0.001).27 The Hawaii study found that a high-intensity CHW intervention in conjunction with a multidisciplinary team was more effective in decreasing mean HgbA1c when compared with usual care with a multidisciplinary team (−2.2 versus 0.2).93 The Hawaii study investigators did not report P value comparing the groups; we were able to calculate it using the data provided in the article and found the difference to be statistically significant (P < 0.0001).93 Project Sugar reported no significant change between the four study groups for the primary outcome, HgbA1c. The only group with a significant improvement from baseline to 2 years was the CHW plus nurse care manager arm (improvement of 0.8 percent ± 0.52 percent, P < 0.05).88–92 Postintervention, a power calculation showed the study was powered to detect a difference of only 1.2 percent change in HgbA1c. Secondary outcomes from Project Sugar included low density lipoprotein (LDL) cholesterol, systolic blood pressure, and diastolic blood pressure; none differed significantly between study groups in change from baseline measures. LDL cholesterol changed for the worse within the CHW plus nurse care manager arm (+4 mg/dl, P < 0.05).88–92
Health care utilization. No study evaluated diabetes care utilization.
Chronic disease management: hypertension. Study characteristics. Four studies (five articles; Table 26), two RCTs23,98,99 and two prospective cohorts,94,95,123 examined outcomes of moderate-intensity CHW interventions for blood pressure management among adult patients with hypertension. We rated one study as fair quality and three as poor quality. All four studies evaluated a CHW intervention compared with an intervention that involved a CHW in a lesser capacity.23,94,95,98,99,123 The two RCTs, one fair98 and one poor99 quality, evaluated CHW interventions in inner-city minority populations.23,98,99
The fair-quality trial from Baltimore, Maryland, evaluated a CHW home visit for patient education, counseling, and referral compared with a CHW home visit plus five additional visits for blood pressure measurement and management, and access to medical care.98 The poor-quality RCT from the West Coast, rated as such because of a high attrition rate, use of a completers analysis, and high potential for bias, evaluated CHW postclinic appointment counseling sessions, CHW home visits, appointment reminder cards and calls, and standard clinical care.23,99
The prospective cohort study from rural central Mississippi, which we rated as poor quality because of a high potential for confounding and inappropriate statistical methods, evaluated a moderate-intensity CHW intervention using CHWs as “hypertension health counselors” in providing monthly visits encouraging compliance with previously prescribed pharmacological and nonpharmacological therapies.94,95 The other prospective cohort study from Baltimore, Maryland, which we rated poor because of a lack of methods describing an analysis plan a priori, a high potential for confounding, and lack of comparison of participant characteristics at baseline, evaluated a moderate-intensity CHW intervention..123 It examined the impact on appointment followup of a CHW followup telephone call after an emergency department visit during which patients had their blood pressure measured, were provided education counseling, and were assisted with appointment keeping and adherence to a treatment plan. The comparison group included patients who had received a single CHW visit in the emergency department but who could not be reached later for assistance in appointment keeping.123 Heterogeneity of study designs, interventions, and outcomes preclude quantitative synthesis of results.
Overview of results. We did not find any fair- or good-quality studies that compared the impact of a CHW intervention with usual care on blood pressure control. Of the three studies that evaluated blood pressure control, only the Mississippi prospective cohort demonstrated a significant difference between study groups in terms of proportion of hypertensive subjects controlled (defined in this study as blood pressure less than 160/95).94,95 Neither RCT demonstrated between-group differences in blood pressure control.23,98,99 However, these studies did note improvement from baseline to study completion within all groups, some of which were statistically significant.23,98,99 The Baltimore prospective cohort did not evaluate blood pressure control but instead examined health care utilization.123 This study demonstrated that CHW worker followup was more effective than no followup in increasing return visit appointment rates.
Knowledge. No study reported improved knowledge.
Behavior. No study reported improved behaviors.
Satisfaction. No study reported satisfaction outcomes.
Health outcomes. We did not find any fair- or good-quality studies that compared the impact of a CHW intervention with usual care on blood pressure control. Three of the four studies did report on blood pressure control. Both RCTs found an improvement within most groups but no difference between groups in terms of blood pressure control.23,98,99 The fair-quality RCT demonstrated that the low-intensity CHW arm (1 home visit) and the high-intensity CHW arm (6 home visits) both improved blood pressure control. However, the difference between the groups was not statistically significant.98 The poor-quality RCT also demonstrated an improvement in blood pressure within all groups, including the usual care arm, but no significant difference between groups.23,99 The Mississippi prospective cohort study did not report statistical tests for either between- or within-group comparisons.94
Health care utilization. The poor-quality prospective cohort in a Baltimore emergency department demonstrated that patients in the low-intensity CHW intervention were more likely to return for a followup appointment than were patients in the comparison group (60 percent versus 40 percent, P < 0.001).123 However, the comparison patients were not able to be contacted for followup by the CHW, thus biasing the results for this outcome in favor of the intervention arm.123
Chronic disease management: infectious diseases. Study characteristics. One RCT of fair quality examined outcomes of a CHW intervention to facilitate access to health care for tuberculosis (TB) in a homeless population with positive purified protein derivative (PPD) test results in San Francisco, California (Table 27).122 This study used a moderate-intensity model. CHWs who were familiar with homelessness were assigned to TB-infected individuals and responsible for accompanying them to their clinic appointments.122 Outcomes were compared with outcomes for a group receiving a monetary incentive to attend the TB clinic in addition to an appointment and bus tokens and with a control group who were given clinic appointments and bus tokens.
Overview of results. This RCT demonstrated that a CHW intervention was less effective than the monetary incentive but more effective than usual care in leading to adherence to a first followup appointment.122
Knowledge. This RCT did not report outcomes for improved knowledge.
Behavior. This RCT did not report outcomes for improved behaviors.
Satisfaction. This RCT did not report outcomes of satisfaction.
Health outcomes. This RCT did not report outcomes of health.
Health care utilization. A moderate-intensity CHW intervention was less effective than a monetary incentive ($5) in increasing adherence to a first followup clinic appointment (75 percent [95% CI, 70–80] versus 84 percent [95% CI, 76–92], P = not reported). However, the CHW intervention was more effective than a control group who received an appointment and bus tokens (75 percent [95% CI, 70–80] versus 53 percent [95% CI, 47–59], P = 0.004).122
Chronic disease management: back pain. Study characteristics. One RCT of fair quality evaluated a moderate-intensity intervention of four 2-hour weekly group classes led by CHWs compared with usual care supplemented by a book on back pain (Table 28).114 The classes focused on applying problem-solving techniques for back pain self-management and included educational materials (book and videos) supporting active management of back pain.114
Overview of results. This fair-quality RCT found that a moderate-intensity CHW intervention was significantly effective in reducing back pain when compared with a control group at 6 months; the groups did not differ significantly at 12 months.114
Knowledge. This RCT did not report outcomes for improved knowledge.
Behavior. This RCT did not report changes in participant behavior.
Satisfaction. This RCT did not report outcomes of satisfaction.
Health outcomes. The moderate-intensity CHW intervention was more effective in decreasing participant back pain than usual care supplemented by a book on back pain at 6 months.114 More participants in the intervention arm achieved a 50 percent or greater reduction in Roland Disability Score from baseline than in the control group at 6 months (47.9 percent versus 33 percent, P = 0.02).114 However, Roland Disability Scores at 12 months did not differ between arms (5.75 ± 6.31 versus 6.75 ±6.39, P = 0.092).114 The authors attributed this lack of difference to the fact that the intervention was intended not to reduce pain intensity but rather to lower patient worries about back pain.114 Additionally, participants receiving a CHW intervention had a lower worry rating (unvalidated tool) than those in the control group at 12 months (2.63 ± 2.58 versus 3.83 ± 3.08, P = 0.013).114
Health care utilization. This RCT did not report on health care utilization.
Other. Participants in the CHW arm reported being more likely to self-manage back or leg pain than those in the control arm, a measure of self-efficacy (77 percent versus 60 percent, P = 0.008).114
Chronic disease management: mental health. Study characteristics. One RCT of poor quality with three trial arms evaluated an assertive community treatment with a CHW intervention compared with an assertive community treatment alone and with a brokered case management intervention (Table 29).120,121 The study population included people in St. Louis, Missouri, who were homeless or at risk for being homeless and were diagnosed with serious psychiatric diagnoses.120,121 The CHWs’ role was to assist with daily living and be available for leisure activities. This intervention was rated as high-intensity as defined in KQ 1. A high rate of attrition (only 85 of 165 provided followup) contributed to the poor-quality rating of this study.120,121
Overview of results. Clients in the assertive community treatment arm plus a CHW did not differ in results when compared with the assertive community treatment group alone, although for many outcomes both of these arms were superior to the brokered case management arm.120,121 The assertive community treatment arms (both with and without a CHW) had more contact with their case managers and were more satisfied than those in the brokered case management arm.120,121 Clients in the assertive community treatment also had fewer psychiatric symptoms at 18 months than clients in the brokered condition.120,121 Days in stable housing did not differ among groups.120,121
Knowledge. This RCT did not report outcomes for improved knowledge.
Behavior. This RCT did not report outcomes for improved behaviors.
Satisfaction. Clients in the assertive community treatment arms (both with and without a CHW) were more satisfied with their treatment program than clients in the brokered case management arm (satisfaction score ± standard deviation: 3.12 ± 0.57 versus 3.27 ± 0.42 versus 2.74 ± 0.68, P < 0.05).120,121
Health outcomes. Clients in the assertive community treatment arm plus a CHW did not differ in health outcome results as compared with the assertive community treatment group alone. Clients in the assertive community treatment arms (both with and without a CHW ) had fewer psychiatric symptoms as rated by the Brief Psychiatric Rating Scale (BPRS) at 18 months compared to baseline than did those in the brokered case management arm (baseline (SD)/18-month followup (SD): 57.97 (20.29)/38.77 (12.23) versus 53.54 (15.54)/39.96 (12.25) versus 50.60 (14.31)/51.60 (16.70), P = 0.001 for any difference among the three groups; P for comparison of either assertive community treatment arm not reported).120,121 Days in stable housing between groups did not differ across the groups.
Health care utilization. Use of health services did not differ between the assertive community treatment plus a CHW arm and the assertive community treatment group alone. Clients in the assertive community treatment arms (both with and without a CHW) had more days in contact with the program than did clients in the brokered case management arm (6.95 (4.91) versus 8.29 (7.51) versus 0.3 (0.49), P < 0.05).
Chronic disease management: asthma. Study characteristics. Two RCTs (three articles), one good-quality,96,97 and one fair-quality,100 examined outcomes of CHW interventions for asthma care among pediatric patients with persistent asthma. Both studies used a highly resource-intensive CHW model. Both studies provided comprehensive multifaceted interventions that included an environmental assessment, asthma action plan, education, referrals, allergy control mattress covers and pillows, vacuums, and cleaning supplies, pest management, and smoking cessation assistance to the high-intensity intervention arm, delivered over a year in several home visits. The Seattle King County Healthy Homes (SKCHH) project (Washington State) compared outcomes for children receiving a high-intensity multivisit home intervention with those for children receiving a low-intensity single home visit that included an environmental assessment, some education, and bedding encasements, followed by the full intervention after a year.96,97 The Community Action Against Asthma (CAAA) project adapted the SKCHH project to Detroit, Michigan, comparing a group receiving the high-intensity multivisit home intervention with a control group receiving an asthma information booklet and the full intervention after a year.100 Variations in measures of health behavior, outcomes, and health care utilization preclude quantitative synthesis of the results.
Overview of results. Two trials demonstrated that high-intensity CHW interventions are more effective than either low-intensity interventions or a control group in reducing unscheduled use of health care services and improving psychological outcomes for caregivers. Both studies demonstrated changes in behavior, such as increased use of bed encasements and vacuuming, associated with the materials distributed by the CHW, but not for other behaviors that may have required external or additional resources or change, such as removal of mold or reduced exposure to environmental tobacco smoke. Both studies demonstrated significant improvements within but not across trial arms for some measures of symptoms,96,97,100 reduced days with activity limitations, and reduced use of beta-agonists.96,97 Authors postulated that these results could be explained either because a minimal intervention may be effective for some outcomes or because of regression to the mean, temporal trends, or the Hawthorne effect (improvement in performance attributable to being observed) among the less intensive or control group participants.96,97 Nevertheless, for health outcomes demonstrating a difference between trial arms such as symptom days, the more intense arm was more effective than the less intense or control arm.
Knowledge. Neither study reported outcomes for improved knowledge of asthma triggers.
Behavior. Both studies examined a variety of behavioral changes (Table 30). Both studies reported increased use of materials provided—that is, mattress covers, pillows, and vacuums, suggesting reduced exposure to dust mites— in the more intense arm. Both studies failed to find differences between the two arms for behavioral changes associated with smoking cessation. Other behaviors that did not differ between arms included removal of pets and use of exhaust fans in the bathroom96,97 and removal of mold.100
Satisfaction. Neither study reported outcomes for satisfaction.
Health outcomes. The SKCHH project reported on the number of symptom days in the past 2 weeks. The CAAA project looked at the occurrence of more than 2 symptom days per week for children not on any controller medication or corticosteroids (Table 31).
Results from these two trials were mixed. The Seattle (SKCHH) project reported nonsignificant differences between the arms in the reduction in symptoms days, whereas the Detroit (CAAA) project found significant differences between the trial arms for children not on any controller medication (OR, 0.39 [95 percent CI, 0.20–0.73]).96,97 The differences between trial arms in reduction of symptom days was not statistically significant in the subset of children not on corticosteroids.100
The Seattle (SKCHH) project also examined differences in trial arms in days with activity limitation, use of beta-agonists, use of controller medications, missed school days for the child, and missed caregiver workdays. With the exception of days with activity limitations, the study found no differences between the intervention arms.96,97 It also found a significantly higher increase in caregiver quality of life (measured by the Center for Epidemiologic Studies Depression Scale in the more intense arm (coefficient for difference between groups in mean change from exit to baseline: 0.58 [95 percent CI, 0.18–0.99]).96,97
The Detroit (CAAA) project found significant improvements in symptoms for both intervention and control arms, but differences were statistically significant only for coughing with exercise and persistent cough. It also found significant differences between trial arms in some but not all measures of lung function; these results could potentially be explained by seasonal influences, changes in instrumentation, and inadequate power.100 Finally, it reported a statistically significant reduction (P = 0.0218) in caregiver depressive symptoms (measured by the Center for Epidemiologic Studies Depression Scale) in the intervention arm (mean value at baseline and followup: 1.62 and 1.54) compared to a rise in depressive symptoms in the control arm (mean value at baseline and followup: 1.58 to 1.64). The study found no statistically significant differences between the two groups in changes in social support between baseline and the endpoint.100
Health care utilization. Both studies (Table 32). found a significant difference in the reduction in unscheduled medical care—emergency room visits, hospitalizations, and unscheduled doctor visits—favoring the more intense intervention at three points: 2 months (OR: 0.38; 95% CI, 0.16–0.89),96,97 3onths (OR: 0.43; 95% CI, 0.23–0.80),100 and 12 months (OR: 0.40; 95% CI, 0.22–0.74).100
KQ 3: Cost-Effectiveness of Community Health Worker Interventions
Overview of Economic Analyses
A total of nine studies that met inclusion criteria for this review contained information about intervention costs, cost-effectiveness, or cost-benefits. We focused here on the six studies that also demonstrated effectiveness of the CHW intervention, either as compared with the alternatives that were analyzed or as compared with baseline, or usual, care.17-22,75,80,96,104,121 The studies on CHWs that included economic information varied a great deal in terms of the populations targeted for intervention, the types of interventions implemented and the settings for those interventions, the alternatives that were analyzed, and the outcomes the interventions sought to impact.
Targeted populations, for example, ranged from Latina women to low-income infants and children. The types of interventions using CHWs as a study arm included early childhood and child health interventions, cancer screening interventions, and chronic disease management interventions. Some studies evaluated alternatives that varied intensity levels for the CHW intervention; others compared the CHW intervention without nurse-delivered interventions; and others compared the CHW intervention with lower intensity alternatives that did not involve direct interaction with targeted patients (e.g., providing written materials only). Study outcomes also varied a great deal across studies, reflecting the diversity of types of interventions and targeted populations (e.g., outcomes related to use of health care, child health and development, and impacts on usual activities such as work or school). Intervention settings also varied; some CHW interventions focused on working with participants in their homes, one focused on working with homeless individuals, and another took place in urban churches.
The three studies with economic information that we eventually excluded involved (1) a diet change intervention that targeted Hispanic women,64 (2) an environmental tobacco smoke intervention that targeted young Latino children,67 and (3) a children’s immunization intervention that compared CHW interventions with mail or telephone interventions for raising children’s immunization rates.69,70 These three studies produced no statistically significant impact on CHW intervention groups as compared with outcomes in the control groups.
In the discussion below, we cite only the articles with data specific to the cost-analysis; studies spanned several other citations specific to outcomes not relevant to the discussion below.
Economics: cancer screening. Study characteristics. Two studies (one trial and one prospective cohort) evaluated program costs or cost-effectiveness for CHW interventions that sought to improve women’s mammography rates.17,22 The ROSE study targeted low-income, rural white, African-American, and Native-American women in North Carolina ages 40 years and older, all of whom had not had a mammogram in the previous 12 months.17,18 These women were randomly assigned to a high-intensity CHW intervention, which involved three home visits with followup telephone calls and mailings, or to a comparison group. The CHW intervention was delivered for a period of 12 to 14 months. The LAMP CHW study collected data on program costs and cost-effectiveness for a low-intensity cancer screening CHW intervention.19–22,104 The intervention was a church-based telephone counseling program that targeted female church members ages 50 to 80 years to promote mammography. Some of these women had obtained mammograms 1 to 2 years before the initial survey and within the 2-year window before that (“adherent” group), whereas others had not (“nonadherent” group). Church volunteers made one telephone call per 12-month period to encourage and address barriers to mammography.
Overview of economic analysis results. Both studies report program costs and the costs per additional mammography screening.17,22 Both studies estimated program costs using a program or funder perspective (i.e., including only those costs that would be incurred by a prevention program to deliver the intervention); that is, they did not employ a societal perspective. Because the LAMP study used volunteer labor, the costs of the intervention from the program perspective are necessarily low compared with the costs of the ROSE intervention, which paid their CHWs. To better understand what costs would be if CHWs had to be hired to deliver the LAMP intervention, Stockdale et al.22 also report two alternative program cost estimates—one that values volunteer time at the minimum wage and another that values volunteer time at the average wage rate.
Measures of effectiveness for economic analysis. The main effectiveness outcome that both studies used for their economic analyses was mammogram receipt in the 12 months before a followup survey. The ROSE study outcomes were based on review of a woman’s medical record.17,22 The LAMP study outcomes were based on participants’ self-reports via a telephone interview; it also estimated life-years saved based on a model of screening, diagnosis, and treatment for breast cancer.22
Economic outcomes. The total cost of the ROSE intervention was estimated to be $329,054,17 which translates to approximately $404 per participant, based on the 815 participants who fully participated in the intervention and data collection. The year of costs was not reported for the ROSE study. Program costs for the LAMP intervention were estimated to be $11 per participant in 1997 dollars when the opportunity cost of CHW volunteers’ time was excluded from the cost calculation. Costs were estimated at $28 per person when CHW volunteers’ time was valued at the minimum wage and $52 per person when an average wage rate for each type of volunteer was used (1997 dollars).22 To compare ROSE and LAMP costs, we assumed that the ROSE costs are in 2000 dollars (the midpoint of the study time period, 1998 through 2002). Using the consumer price index for all urban consumers (CPI-U) to adjust the LAMP intervention cost of $52 per participant in 1997 dollars to 2000 dollars yields an estimate of $56 per LAMP participant, as contrasted with the high-intensity ROSE intervention cost of about $404 per participant.
Both studies also reported costs per additional screening.17,22 Paskett et al. estimated the impact of the ROSE intervention to be 66 additional mammograms in the CHW intervention group, resulting in a cost-effectiveness ratio of $4,986 per additional mammogram (assumed 2000 dollars).17 Stockdale et al. estimated the impact of the LAMP intervention to be 3.24 additional mammography screenings for each of the 45 churches that participated in the study, resulting in an estimated cost per additional screening of $903 (1997 dollars) when CHW volunteers’ time was valued at the average wage rate.22 Although these findings appear to suggest that the LAMP intervention had a much lower cost per additional mammogram received than did the ROSE intervention, the effectiveness and cost-effectiveness results are not comparable between these two studies because the LAMP intervention targeted women who were both adherent and nonadherent with screening guidelines, whereas the ROSE intervention targeted only nonadherent women. Focusing on results for the nonadherent LAMP participants only, the estimated intervention effectiveness is 1.46 additional screenings per church per year (not statistically significant), which we estimate to produce a cost-effectiveness ratio of $2,005 per additional mammography screening in 1997 dollars, or $2,151 in 2000 dollars, when the time of CHWs is valued using expected wage rates.
Stockdale et al. also estimated the cost per life-year saved by the LAMP intervention and subsequent mammography screening as $46,308 (1997 dollars),22 when CHW time was valued using expected wage rates ($33,632 plus the estimated cost per life-year saved for mammography screening of $12,676).
Economics: chronic disease management. Study characteristics. Two studies provided economic information on the management of chronic diseases; both studies are described in more detail in the last section of KQ 2. One study evaluated an asthma control intervention for children;96 the other evaluated an intervention to prevent homelessness in patients with mental illness.121
The asthma intervention, known as the Seattle King County Healthy Homes (SKCHH) project, evaluated a 1-year high-intensity CHW intervention approach, involving five to nine CHW home visits.96 The investigators compared this high-intensity intervention with a low-intensity version that involved only one CHW home visit and evaluated health care utilization and costs for participants, intervention program costs, and other measures related to asthma control, quality of life, and productivity.
The homelessness prevention intervention compared three alternative case management approaches for people with mental illness at high risk of homelessness:
- brokered case management—a low-intensity intervention that can be viewed as the baseline, or usual care, approach;
- assertive community treatment—a high-intensity intervention that involves frequent interaction with the client and assistance with a host of activities and social service acquisition; and
- assertive community treatment with CHWs—a high-intensity CHW intervention that consists of assertive community treatment, adding a CHW to interact with and assist clients.121
Each intervention was provided over 18 months. Key outcomes were health care and social services utilization, program costs, and pre- and postintervention measures of 6 months of costs for health care and social services among study participants.
Overview of economic analysis results. Both studies report program costs per participant from the program perspective. For the Seattle study, Krieger et al.96 estimated the cost of the 12-month intervention by summing payments for salary and fringe benefits, supplies, rent, travel, and office expenses and adding indirect costs of 13 percent. For the homelessness prevention study, Wolff et al.121 estimated the additional intervention program costs of assertive community treatment, with and without CHWs, as the costs above those for brokered case management; their estimates values CHW time at the minimum wage.
Both studies estimated the impact of the intervention on health care and/or social services costs for program participants. For example, Krieger et al. assessed the pre- and postintervention costs of urgent care services for both CHW intervention arms (high and low intensity).96 Wolff et al. also assessed the pre- and postintervention costs of the following services for program participants in all three intervention arms: mental and physical health, vocational and educational, residential, and supportive social.121
Measures of effectiveness for economic analysis. Neither study created a measure of the costs per unit of program effectiveness (e.g., cost per additional day in stable housing or cost per additional day of school attendance). Instead, both studies estimated program cost savings or potential cost savings by comparing health care or social services costs in the preintervention time period with costs in the postintervention time period.
For example, the Seattle study estimated urgent care costs for the targeted children in the 2 months before the start of the intervention and compared these values without analogous costs in the 2-month period before the exit interview.96 For this work, Krieger et al. defined urgent care costs as the costs of hospital admissions, emergency department visits, and unscheduled clinic visits.96 Because this intervention sought to reduce use of urgent health care services among participants with asthma, a reduction in urgent care costs for participants may be viewed as cost savings attributable to the intervention.
The homelessness prevention intervention also compared preintervention and postintervention costs for participants in each of the study arms. For this work, Wolff et al. calculated costs for the following services, by study arm:
- mental health inpatient,
- mental health outpatient,
- physical health inpatient,
- physical health outpatient,
- vocational and educational,
- cash social support, and
- in-kind social support.121
Wolff et al. also provided a total cost amount that summed the per-patient costs for all of the above services and included the intervention cost for assertive community treatment (with or without CHWs).121 However, reductions in these total or specific services costs should not be viewed as cost savings attributable to the intervention because utilization and costs of some services might be expected to rise, rather than fall, as the result of a successful intervention. For example, successful assertive community treatment interventions might lead to larger pre/postintervention increases in vocational and educational service costs than a brokered case management approach.
Economic outcomes. In the Seattle program, costs for the high-intensity CHW intervention were $1,124 per child higher in 2001 dollars than costs for the low-intensity CHW intervention.96 Estimated costs for the low-intensity asthma intervention were not provided.96
For the homelessness prevention intervention, annual program costs were $6,200 per participant for the assertive community treatment intervention with CHWs and $6,440 per participant for assertive community treatment only.121 These cost estimates are in 1992 dollars and are in addition to costs for brokered case management—costs that were not reported in the article. Adjusting these cost estimates to 2001 dollars using the CPI-U, we estimate the costs of assertive community treatment with CHWs to be $7,826 per patient and the costs of assertive community treatment only to be $8,129 per patient, in addition to costs for brokered case management.
For the Seattle study of children with asthma, Krieger et al. also provided estimates of pre/postintervention health care cost reductions attributable to the CHW asthma intervention.96 Comparing urgent care costs in the 2 months before the intervention with costs in the 2 months at the end of the intervention, they estimated cost reductions of $201 to $334 per child in 2001 dollars.96 For the low-intensity CHW group, analogous cost reductions were $185 to $315 per child. Assuming these cost reductions persist for 1 year, estimated annual cost reductions are $1,200 to $2,000 per child for the high-intensity CHW intervention in 2001 dollars. Krieger et al. also discussed the cost-effectiveness of the high-intensity intervention relative to the low-intensity approach.96 They found savings in urgent care costs for the high intervention group relative to the low intervention group of $57 to $80 per child over a 2-month period.96 The authors reported that if these cost reductions were to last for 3 to 4 years, the high-intensity intervention would be cost saving relative to the low-intensity intervention. Whether assuming the same level of reduced urgent care utilization and costs for several years postintervention is reasonable, however, remains unclear. The authors did find that urgent care utilization remained low in the high-intensity group for at least 6 months following the intervention.96
For the study of homeless mentally ill participants, Wolff et al. conducted regression analyses to explore whether study arms differed in their measures of total costs over the 18-month study period.121 They found no difference in total costs across study arms after controlling for patients’ costs in the preintervention period.121 They also compared 6-month costs in the preintervention period to 6-month costs for three separate postintervention periods (1 to 6 months, 7 to 12 months, and 13 to 18 months). At least in part because the number of participants in each intervention arm was relatively small (N = 35, 28, and 22, respectively), postintervention costs varied a great deal across time periods. The authors point out that, when comparing the preintervention period with the first 6 months’ postintervention period, inpatient mental health services costs fell $1,315 for assertive community treatment with CHWs, rose almost $4,500 for assertive community treatment only, and rose more than $8,000 for brokered case management.121 Considering the second 6-month period of the intervention, inpatient mental health services costs fell by more ($4,400 per participant) in the assertive community treatment only group than in the assertive community treatment with CHWs group ($2,651 per participant). Inpatient mental health services costs also declined in that time period for the brokered case management group ($1,252 per participant). All of these cost estimates are in 1992 dollars.
The cost estimates for health and social services that Wolff et al.121 report are difficult to interpret using the currently recommended framework for performing and evaluating cost-effectiveness analyses.130,131 The recommended approach for performing cost-effectiveness analysis is to specify the perspective of the study a priori and to calculate net costs for use in cost-effectiveness evaluation as intervention costs less any costs for health care or other relevant variables (including productivity losses) averted by the intervention. The societal perspective is recommended for economic evaluation, which implies that all costs should be included, regardless of who bears them. Although the Wolff et al.121 article implies that the intent was to estimate costs from the societal perspective, their measure of total costs excludes criminal justice and family burden costs (mentioned as a limitation), it excludes productivity costs, and it includes societal transfers (cash and in-kind support) that are not recommended for inclusion in economic analyses from the societal perspective. The presentation of costs in three different intervention time periods also makes it difficult to interpret the Wolff et al. estimates, because costs differed a great deal over time for each intervention arm.121 Finally, the total cost measures they presented cannot readily be used in economic evaluations without some adjustments. Their total cost estimates represent the sum of intervention costs and specific health care and social services costs. These total cost estimates vary a great deal across intervention arms (including for the preintervention period) and across time within each intervention arm. In contrast, the recommended estimates for use in economic evaluations are measures of net costs that provide a single measure of costs for each intervention arm that subtract from intervention costs the health care, productivity, and other related cost reductions attributable to the intervention.131
Economics: child health. Study characteristics. Two studies evaluated program costs for CHW interventions that sought to improve child health.75,80 One study, set in Maryland, evaluated the impact of a high-intensity CHW intervention for children with nonorganic failure to thrive in a low-income urban setting. As reported by Black et al., children diagnosed with failure to thrive were randomized to receive either the CHW intervention, which involved the delivery of clinical services plus weekly home visits from a trained CHW, or the clinical intervention only.75 The Home Visitation 2000 RCT targeted low-income, pregnant women for a home visiting intervention that involved prenatal home visits, followed by home visits every 1 to 2 months until the target child was 2 years of age.80 In this study, Olds et al. compared the impact of using CHWs to deliver the home visiting intervention with the impact of using nurses.80 In addition to program costs, it evaluated several child health and developmental outcomes (e.g., mother-child interaction, quality of the home environment, child developmental outcomes).
Overview of economic analysis results. Both studies reported intervention program costs. Cost components for the failure-to-thrive trial included salaries for the CHW or nurse, materials costs, transportation costs, costs of police service, and a 10 percent overhead fee.75 Olds et al. provided a per-family total cost of the 2.5 year Home Visitation 2000 trial,80 but they did not specify details on what was included in the cost estimate.
Measures of effectiveness for economic analysis. Because both of these studies reported only the intervention costs, they did not examine intervention costs in relationship to outcomes. Thus we had no measures of intervention effectiveness for these economic analyses.
Economic outcomes. Annual program costs for the failure-to-thrive CHW intervention were $2,828 per child in 1993 dollars.75 Although the article did not explicitly state this, we assumed that this cost estimate reflects the additional costs of the CHW intervention relative to the clinical intervention (usual care). When adjusted to 2002 dollars using the CPI-U, the CHW intervention for these children has an estimated annual cost of $3,520 per child. For the Home Visitation 2000 RCT, program costs were $9,140 per family in 2002 dollars for the nurse home visitation arm and $6,162 per family for the CHW intervention arm.80 These costs are for the full 2.5 years of the program. Dividing these estimates by 2.5, we estimate annual costs of $3,656 per family for the nurse home visitation intervention and $2,465 for the CHW home visitation intervention—both in 2002 dollars.
Economics: summary of findings. Table 33 summarizes findings from the six CHW intervention articles that provided information on program costs and other economic outcomes (presented in the order of discussion above, by clinical context). Cost estimates are shown as presented in each article, but we also report each cost estimate adjusted to 2008 dollars using the CPI-U. Although adjusting some of the cost estimates using the medical care component of the CPI might have been more appropriate, because that component accounts for faster growth in prices in the health care sector than in other parts of the US economy, we used the CPI-U because all studies relied on nonmedical labor to provide the CHW intervention.
KQ 4: Training of Community Health Workers
Characteristics of Training for Community Health Workers
Overview. Study characteristics. As noted in Chapter 2, an inclusion criterion specific to KQ 4 was that all studies reported on changes in knowledge or skills among CHWs after training. Although we identified 46 citations that were potential includes,111,132–176 only 9 studies (10 citations) provided evidence of changes in knowledge or skills among CHWs after training.137,141,143,147–150,155,169,176
All included studies were set in minority or underserved communities. Three focused on cancer prevention,137,141,143,176 two on cardiovascular disease,147,149 and one each on health promotion,169 tobacco cessation,150 salmonella prevention in the manufacture of queso fresco,155 and on health insurance enrollment, immunizations, and asthma prevention.148
The studies included in this section spanned a variety of models of CHW interventions. Five studies relied on volunteers;137,141,149,155,169,176 other studies either paid CHWs or did not report on payment status. The size of the intervention effort also varied: the number of CHWs trained through these programs ranged from 4147 to 1,504.148 The educational background and prior training of the CHWs undergoing training were rarely reported: one study reported that 98 percent (of 79 CHWs) had either a college bachelor’s or graduate degree,137 whereas another study reported that all trainees (4 CHWs) had 10 years of prior experience as CHWs.147 Studies also varied in their degree of specificity in reporting eligibility criteria for CHWs. The contribution of CHWs to developing training materials varied, ranging from intensive involvement in pretesting to no involvement. Studies also varied in their reporting on training components; in the following sections, we describe reported data on components of training.
Training on recruitment and retention process skills. Two studies reported training on recruitment and retention.141,150,176 One study noted that client recruitment was addressed, but the content, method, and number of sessions was not reported.150 The other recorded five 2-hour sessions covering recruitment strategies and role-playing practice.141,176
Training on intake and assessment. One study reported training for intake and assessment, specifically on community mobilization, communication skills, and outreach strategies, but it provided no details on the content, method, and number of sessions.148 A second study noted two training sessions for assessment and role-play.147
Training on protocol delivery. Two studies reported on training on protocol delivery.147,148 One provided no further details,148 and the second listed health education counseling as part of the curriculum, and included role play for cancer screening counseling sessions and cardiovascular disease counseling sessions that was followed by external feedback from a clinical psychologist.147
Training on health topic. The purpose of training CHWs on health topics was to prepare them to educate participants. Seven studies described the health content of their training in some detail;137,141,143,147,150,155,169,176 all provided evidence of change in knowledge of skills after training (Table 34). Only two reported significance tests.
Training on evaluation. A single study reported evaluation as one of the seven core modules in their curriculum but provided no further details.148
Other training. Four studies reported training on communication skills,137,148,149,176 and a single study reported on making referrals.176 Training curricula may well have included additional elements that were not reported.
Patient Outcomes of Community Health Worker Training
We did not identify any studies that reported on patient health outcomes of CHW interventions that were linked to characteristics of training.
Agency for Healthcare Research and Quality (US), Rockville (MD)
Viswanathan M, Kraschnewski J, Nishikawa B, et al. Outcomes of Community Health Worker Interventions. Rockville (MD): Agency for Healthcare Research and Quality (US); 2009 Jun. (Evidence Reports/Technology Assessments, No. 181.) 3, Results.