Abstract
Objective
To compare utilization and preventive care receipt among patients of federal Section 330 health centers (HCs) versus patients of other settings.
Data Sources
A nationally representative sample of adults from the Medical Expenditure Panel Survey (2004–2008).
Study Design
HC patients were defined as those with ≥50 percent of outpatient visits at HCs in the first panel year. Outcomes included utilization and preventive care receipt from the second panel year. We used negative binomial and logistic regression models with propensity score adjustment for confounding differences between HC and non-HC patients.
Principal Findings
Compared to non-HC patients, HC patients had fewer office visits (adjusted incidence rate ratio [aIRR], 0.63) and hospitalizations (aIRR, 0.43) (both p < .001). HC patients were more likely to receive breast cancer screening than non-HC patients (adjusted odds ratio [aOR] 2.78, p < .01). In subgroup analyses, uninsured HC patients had fewer outpatient and emergency room visits and were more likely to receive dietary advice and breast cancer screening compared to non-HC patients.
Conclusions
Health centers add value to the health care system by providing socially and medically disadvantaged patients with care that results in lower utilization and maintained or improved preventive care.
Keywords: Safety net, preventive care, utilization
Health centers (HCs) funded by Section 330 of the Public Health Service Act cared for approximately 21.1 million people in 2012. This program is dedicated to providing comprehensive primary care to medically vulnerable populations. The number of patients cared for at HCs is expected to grow considerably due to increases in program funding and the expansion of the insured population, resulting from implementation of the Affordable Care Act.
It is important to understand what additional value HCs provide to the health care system beyond the care provided by other primary care sites. HCs provide comprehensive primary care to patients, regardless of their ability to pay for care. Comprehensive primary care includes a unique array of enabling services (e.g., case management, translation, and transportation) and may include selected dental, behavioral health, pharmacy, and other services not typically available in primary care settings (Shi et al. 2010). HCs must be located in or serve medically underserved areas/populations, which are areas having too few primary care providers, high infant mortality, high poverty, and/or a large elderly population.
Previous studies examining the impact of HCs on health care utilization have produced mixed results. Some studies have suggested that receipt of primary care at HCs decreases emergency room (ER) visits (Smith-Campbell 2005; Rust et al. 2009) and hospitalizations (Epstein 2001; Probst, Laditka, and Laditka 2009; Rothkopf et al. 2011). Among Medicaid beneficiaries, the benefit of primary care at HCs has been associated with fewer hospitalizations and ER visits for ambulatory care–sensitive conditions (Falik et al. 2001, 2006). However, other studies have found that primary care at HCs was associated with no difference in preventable hospitalizations (Gurewich et al. 2011), and at worst, increased ER visits compared to other primary care sites (Scherer and Lewis 2010). The disparate results are likely because previous studies have evaluated different populations and used various data sources. Further, while inpatient and ER visits have been extensively examined, the study of the effects of HCs on overall outpatient health care utilization has been limited.
Studies on the quality of care at HCs have been generally positive (Starfield et al. 1994; Falik et al. 2001, 2006; Porterfield and Kinsinger 2002; Regan et al. 2003; Hicks et al. 2006; Shi and Stevens 2007; Goldman et al. 2012). In the outpatient setting, HCs have performed better on some, but not all, measures of quality compared to other primary care providers (Porterfield and Kinsinger 2002; Regan et al. 2003; Hicks et al. 2006; Shi and Stevens 2007; Goldman et al. 2012).
Studying the effects of HCs on utilization and quality of care has unique challenges. First, few nationally representative datasets of health care providers provide accurate identification of HCs, which leads to inaccurate identification of HC patients. Data on providers reported by patients may be inaccurate, as patients may not know if their primary care clinic is a federally funded HC or another type of community health center. Second, due to the HC mission, HC patients are more likely to be racial/ethnic minorities, uninsured or Medicaid enrollees, and have higher rates of chronic disease (Forrest and Whelan 2000; Shi et al. 2010). These important differences may be a source of confounding if not accounted for in analyses. Third, HCs are intentionally located in federally designated medically underserved areas. Thus, HC patients may be located in different geographic locations than other primary care patients and have different access to health care resources. These differences in location, which may affect utilization, have not been accounted for in previous analyses.
Previous studies on HCs have left uncertainty about the effects of this safety net program on health care utilization and quality of care. We hypothesized that patients receiving care at HCs would have no difference in outpatient visits, but fewer emergency room visits and hospitalizations, and similar or better quality of care compared to similar patients receiving care at other sites. In this study, we have addressed some limitations of past work by address-matching federally funded HCs to provider addresses, including less readily available variables in adjustment for selection (e.g., quality of life), using advanced methods of adjustment (propensity scores), and adjusting for differences in patient proximity to HCs.
Methods
We used five panels from the Medical Expenditure Panel Survey (MEPS) (2004–2008) (N = 79,041), which is a nationally representative household survey of the noninstitutionalized civilian population performed by the Agency for Healthcare Research and Quality (AHRQ). MEPS is a 2-year panel survey and collects data in five household interviews. We also used Health Resources and Services Administration Bureau of Primary Health Care (BPHC) Uniform Data System and BPHC Management Information System databases. These administrative databases include addresses of all Section 330 HCs and affiliated sites from 2004 to 2008. This project was approved by institutional review boards at the University of Chicago and NORC at the University of Chicago.
We studied adults, aged ≥18 years, with ≥1 clinic (office or hospital-based) visit in their first panel year, and who lived ≤20 miles of an HC (n = 33,137). We used data from the first panel year to identify participants to study how the site of care may affect future health care utilization. In MEPS, participants are queried at each interview whether their household used health care services and the address of each provider (Richard et al. 2012). To accurately identify visits at HCs, we compared provider addresses in MEPS to HC addresses in BPHC databases. Congruent with previous studies, we considered participants to be HC patients if the majority (≥50 percent) of their clinic visits were at HCs (Falik et al. 2001; Gurewich et al. 2012); other participants were considered non-HC patients. Because patients tend to live near their site of health care, and differences in proximity may affect their utilization of their health care site (Hadley and Cunningham 2004; Gresenz, Rogowski, and Escarce 2007), we restricted the sample to participants living ≤20 miles of an HC. Details on how this distance was calculated are available in the online-only supplement (Data S1).
We included the MEPS year-end summary data from the first panel year on sociodemographics, health behaviors, comorbid illnesses, quality of life (Ware, Kosinski, and Keller 1996), depressed symptoms (Gilbody et al. 2007), and household geography (see Table 1 for complete variable list). We constructed variables to describe each participant’s type(s) of insurance (Medicare, Medicaid, private, Medicaid HMO, private HMO) and whether they held that insurance for the full year, part of the year, or not at all. Participants who had no health insurance for the entire first panel year were considered uninsured. We also counted the number of functional limitations with which participants reported having some difficulty (lifting 10 pounds, walking up 10 steps, walking three blocks, walking a mile, standing 20 minutes, bending/stooping, reaching overhead, using fingers to grasp). In addition, we used U.S. Census zip code data from 2007 to 2009 on the poverty rate and proportion of minority race and Hispanics. Lastly, we included the 2006 county-level urban–rural classification from the National Center of Health Statistics.
Table 1.
Adult Participants Living within Twenty Miles of a Health Center Receiving Outpatient Care in the First Panel Year (N = 33,137)
| Health Center (n = 976*) | Non–Health Center (n = 32,161) | p-Value | |
|---|---|---|---|
| Demographics | |||
| Age, years, mean (SE) | 42.3 (0.8) | 48.3 (0.2) | <.001 |
| Female, n (%) | 689 (64) | 19,275 (58) | .007 |
| Marital status, n (%) | |||
| Married | 467 (43) | 18,403 (57) | <.001 |
| Widowed | 60 (6) | 2,801 (9) | |
| Divorced | 109 (13) | 3,809 (12) | |
| Separated | 59 (5) | 830 (2) | |
| Never married | 281 (33) | 6,317 (20) | |
| Race/ethnicity, n (%) | |||
| White | 178 (33) | 18,222 (73) | <.001 |
| Black | 201 (24) | 5,298 (11) | |
| Hispanic | 554 (38) | 6,243 (11) | |
| American Indian/Alaskan Native | 3 (0.6) | 143 (0.5) | |
| Asian | 23 (3) | 1,698 (4) | |
| Hawaiian/Pacific Islander | 3 (0.3) | 113 (0.3) | |
| Other | 14 (2) | 444 (1) | |
| Education, years, mean (SE) | 11.1 (0.2) | 13.3 (0.03) | <.001 |
| Percent federal poverty line, n (%) | |||
| <100% | 404 (34) | 4,725 (10) | <.001 |
| 100–124% | 78 (8) | 1,743 (4) | |
| 125–199% | 225 (23) | 4,626 (12) | |
| 200–399% | 201 (23) | 9,464 (30) | |
| ≥400% | 68 (12) | 11,603 (43) | |
| Language most often spoken at home, n (%) | |||
| English | 508 (69) | 27,016 (91) | <.001 |
| Spanish | 434 (28) | 3,702 (5) | |
| Other | 29 (3) | 1,207 (3) | |
| Insurance status | |||
| Private insurance, n (%) | |||
| Full year | 133 (19) | 18,139 (63) | <.001 |
| Partial year | 72 (9) | 2,983 (10) | |
| Private HMO insurance, n (%) | |||
| Full year | 45 (6) | 5,869 (19) | <.001 |
| Partial year | 48 (5) | 2,461 (8) | |
| Medicaid insurance, n (%) | |||
| Full year | 230 (20) | 2,860 (5) | <.001 |
| Partial year | 142 (13) | 1,819 (4) | |
| Medicaid HMO insurance, n (%) | |||
| Full year | 132 (12) | 1,424 (3) | <.001 |
| Partial year | 92 (8) | 1,182 (3) | |
| Medicare insurance, n (%) | |||
| Full year | 121 (15) | 7,042 (21) | <.001 |
| Partial year | 13 (1) | 610 (2) | |
| Health behaviors | |||
| Has usual source of care, n (%) | 800 (79) | 26,915 (84) | .009 |
| Physical activity ≥3 times per week, n (%) | 429 (47) | 16,630 (55) | .002 |
| Current smoker, n (%) | 191 (24) | 5,330 (16) | <.001 |
| Comorbid conditions | |||
| Hypertension, n (%) | 348 (36) | 11,627 (34) | .55 |
| Heart disease, n (%) | 32 (6) | 1,693 (5) | .63 |
| Angina, n (%) | 27 (3) | 980 (3) | .81 |
| Myocardial infarction, n (%) | 35 (6) | 1,309 (4) | .27 |
| Stroke, n (%) | 32 (4) | 1,273 (4) | .73 |
| Emphysema, n (%) | 12 (2) | 687 (2) | .67 |
| Diabetes, n (%) | 155 (14) | 3,913 (10) | .004 |
| Arthritis, n (%) | 202 (22) | 8,688 (27) | .08 |
| Asthma, n (%) | 97 (11) | 3,547 (11) | .79 |
| Obesity, n (%) | 350 (35) | 9,652 (28) | <.001 |
| Depression score,† range, 0 (best)–6 (worst), mean (SE) | 1.2 (0.08) | 0.8 (0.01) | <.001 |
| Number functional limitations, range, 0–8, mean (SE) | 1.2 (0.12) | 1.1 (0.02) | <.001 |
| Health-related quality of life | |||
| Mental HRQL,‡ range, 0 (worst)–100 (best), mean (SE) | 47.4 (0.5) | 50.2 (0.1) | <.001 |
| Physical HRQL,‡ range, 0 (worst)–100 (best), mean (SE) | 47.4 (0.5) | 48.4 (0.1) | <.001 |
| Household geography | |||
| Distance to nearest health center, miles, mean (SE) | 2.8 (0.2) | 6.0 (0.1) | <.001 |
| Household in a metropolitan statistical area, n (%) | 823 (83) | 28,008 (87) | .19 |
| Zip code-level characteristics | |||
| Poverty rate in zip code, n (%) | 20.1 (0.6) | 13.3 (0.2) | <.001 |
| Proportion of minority race in zip code, n (%) | 40.2 (1.6) | 27.3 (0.5) | <.001 |
| Proportion of Hispanics in zip code, n (%) | 26.7 (0.2) | 14.4 (0.4) | <.001 |
| County-level urban–rural classification, n (%) | |||
| Large metropolitan, central | 406 (38) | 11,613 (32) | <.001 |
| Large metropolitan, fringe | 99 (13) | 7,495 (26) | |
| Medium metropolitan | 217 (19) | 6,497 (21) | |
| Small metropolitan | 112 (15) | 2,641 (9) | |
| Micropolitan | 66 (6) | 2,501 (8) | |
| Counties not in a metropolitan statistical area | 76 (9) | 1,414 (4) | |
HC patients were those who received the majority (≥50%) of office-based health care in an HC during the first panel year of MEPS.
Depression score was measured using the Patient Health Questionnaire-2.
Health-related quality of life was measured using the SF12-v2.
HC, health center; HMO, health management organization; HRQL, health-related quality life; MEPS, Medical Expenditure Panel Survey; SE, standard error.
The outcomes of interest were patient-reported health care utilization and preventive care in the second panel year. Health care utilization was defined by total office visits, hospital-based outpatient visits, prescriptions filled, ER visits, and hospitalizations in the second panel year. Preventive care included the following: dietary/exercise advice in the last 2 years, influenza vaccination in the last year, hypertension screening in the last year, hyperlipidemia screening ever (men aged ≥35 years; women aged ≥45 years), cervical cancer screening in the last 3 years (women aged 21–65 years) (Moyer 2012), breast cancer screening in the last 2 years (women aged 50–74 years) (2009), and colon cancer screening ever (adults aged 50–75 years) (2008).
Statistical Methods
We evaluated differences in utilization and receipt of preventive care services using bivariate (referred to as unadjusted) and multivariate (adjusted) statistical techniques. Differences between HC and non-HC patients were evaluated with χ2 tests for categorical and t-tests for continuous variables. Because the characteristics of HC patients are dissimilar from other primary care patients (Forrest and Whelan 2000; Shi et al. 2010), we used propensity score methods to balance potential confounders (Rosenbaum and Rubin 1983; Rubin 1997; D’Agostino 1998). Propensity scores were estimated using logistic regression based on 37 variables (see Table 2, footnote §). Missing values for covariates were imputed using the weighted Hot-Deck imputation method prior to propensity score estimation (Data S1) (Reilly and Pepe 1997). The propensity score analysis was informed by computing the standardized bias for each variable (Harder, Stuart, and Anthony 2010). Standardized bias values <0.25 were considered well balanced (Data S1 and Table S1).
Table 2.
Health Care Utilization and Receipt of Preventive Care among Adult MEPS Participants (N = 33,137†)
| Health Center (n = 976†) | Non–Health Center (n = 32,161†) | Unadjusted IRR or OR (95% CI) | Adjusted IRR or OR (95% CI)‡,§ | |
|---|---|---|---|---|
| Health care utilization | ||||
| Office visits, mean (SE) | 4.67 (0.36) | 7.57 (0.09) | 0.62 (0.53–0.72)*** | 0.63 (0.52–0.76)*** |
| Hospital outpatient visits, mean (SE) | 0.45 (0.11) | 0.70 (0.03) | 0.64 (0.40–1.03) | 0.98 (0.42–2.31) |
| Prescription drugs, mean (SE) | 15.5 (1.13) | 16.5 (0.22) | 0.94 (0.82–1.08) | 0.91 (0.77–1.06) |
| Emergency room visits, mean (SE) | 0.31 (0.04) | 0.21 (0.004) | 1.46 (1.14–1.88)** | 1.18 (0.79–1.75) |
| Hospitalizations, mean (SE) | 0.12 (0.01) | 0.15 (0.004) | 0.81 (0.64–1.01) | 0.43 (0.30–0.61)*** |
| Receipt of preventive care | ||||
| Diet advice, n (%) | 423 (42) | 13,213 (41) | 1.05 (0.87–1.26) | 1.30 (0.93–1.80) |
| Exercise advice, n (%) | 456 (46) | 14,572 (46) | 1.00 (0.83–1.21) | 1.21 (0.86–1.70) |
| Influenza vaccination, n (%) | 284 (30) | 12,340 (41) | 0.60 (0.49–0.73)*** | 0.83 (0.59–1.17) |
| Hypertension screening, n (%) | 817 (87) | 27,337 (89) | 0.85 (0.64–1.12) | 1.33 (0.83–2.12) |
| Hyperlipidemia screening, n (%) | 402 (94) | 18,399 (97) | 0.58 (0.36–0.94)* | 1.34 (0.62–2.87) |
| Cervical cancer screening, n (%) | 522 (92) | 12,707 (89) | 1.33 (0.93–1.91) | 1.32 (0.70–2.50) |
| Breast cancer screening, n (%) | 141 (84) | 5,213 (83) | 1.04 (0.62–1.75) | 2.78 (1.49–5.18)** |
| Colon cancer screening, n (%) | 93 (49) | 6,044 (70) | 0.42 (0.29–0.60)*** | 0.59 (0.33–1.03) |
Number of participants could be smaller due to missing outcomes.
Results are adjusted for propensity scores applied as inverse probability treatment weights, private/Medicaid insurance status, private/Medicaid HMO insurance status, poverty status, census region, MEPS panel year, and variables with standardized biases >0.25 (marital status, Medicare insurance, percent of federal poverty line, number of functional limitations, and urban–rural continuum code) after propensity score adjustment.
Variables in propensity score estimation include the following: age, gender, marital status, race/ethnicity, education, private/Medicare/Medicaid insurance, private/Medicaid HMO insurance, percent of federal poverty line, language most often spoken at home, having a usual source of care, physical activity, smoking status, hypertension, heart disease, angina, myocardial infarction, stroke, emphysema, diabetes, arthritis, asthma, obesity, mental and physical health–related quality of life, depressed symptoms, number of functional limitations, household in a metropolitan statistical area, distance to the nearest health center, zip code–level poverty rate, zip code–level proportion of minority race, zip code–level proportion of Hispanics, urban–rural continuum code, MEPS survey year, and MEPS survey person weight variable.
***p < .001, **p < .01, *p < .05.
HMO, health management organization; IRR, incidence rate ratio; MEPS, Medical Expenditure Panel Survey; SE, standard error.
All analyses included MEPS survey design variables and weights. In adjusted analyses, survey weights were included after being multiplied by inverse probability treatment weights from propensity scores (Zanutto, Lu, and Hornik 2005; Zanutto 2006; Stuart 2010; Dugoff, Schuler, and Stuart 2013). Variables with standardized biases ≥0.25 were included in subsequent regressions to account for any remaining potential influence. Other covariates in subsequent regressions included characteristics from the second panel year that may have changed since the first panel year (private/Medicaid insurance status, private/Medicaid HMO insurance status, and poverty status), census region, and MEPS panel year (to account for trends) (see Tables 2 and 3, footnote ‡). In adjusted analyses, health care utilization was modeled using negative binomial regression, and receipt of preventive care was modeled using logistic regression.
Table 3.
Health Care Utilization and Receipt of Preventive Care among Full-Year Uninsured Adult MEPS Participants (N = 4,090†)
| Health Center (n = 359†) | Non–Health Center (n = 3,731†) | Unadjusted IRR or OR (95% CI) | Adjusted IRR or OR (95% CI)‡,§ | |
|---|---|---|---|---|
| Health care utilization | ||||
| Office visits, mean (SE) | 2.92 (0.32) | 4.53 (0.19) | 0.65 (0.51–0.82)*** | 0.70 (0.51–0.96)* |
| Hospital outpatient visits, mean (SE) | 0.25 (0.06) | 0.43 (0.05) | 0.58 (0.34–1.01) | 0.51 (0.28–0.91)* |
| Prescription drugs, mean (SE) | 9.70 (1.28) | 9.42 (0.35) | 1.03 (0.79–1.35) | 0.80 (0.60–1.07) |
| Emergency room visits, mean (SE) | 0.21 (0.04) | 0.23 (0.01) | 0.91 (0.63–1.32) | 0.59 (0.37–0.95)* |
| Hospitalizations, mean (SE) | 0.10 (0.02) | 0.08 (0.01) | 1.26 (0.83–1.90) | 0.82 (0.49–1.37) |
| Receipt of preventive care | ||||
| Diet advice, n (%) | 162 (44) | 1,267 (32) | 1.72 (1.28–2.32)*** | 1.98 (1.29–3.04)** |
| Exercise advice, n (%) | 172 (48) | 1,470 (37) | 1.53 (1.13–2.07)** | 1.43 (0.93–2.21) |
| Influenza vaccination, n (%) | 78 (22) | 844 (23) | 0.93 (0.64–1.37) | 0.72 (0.44–1.16) |
| Hypertension screening, n (%) | 290 (83) | 2,785 (77) | 1.47 (0.98–2.22) | 1.65 (0.82–3.31) |
| Hyperlipidemia screening, n (%) | 124 (93) | 1,587 (90) | 1.49 (0.74–2.98) | 2.49 (0.66–9.37) |
| Cervical cancer screening, n (%) | 212 (90) | 1,630 (81) | 2.07 (1.25–3.41)** | 1.57 (0.69–3.57) |
| Breast cancer screening, n (%) | 52 (87) | 387 (68) | 3.23 (1.44–7.25)** | 4.53 (1.65–12.4)** |
| Colon cancer screening, n (%) | 29 (50) | 339 (49) | 1.07 (0.57–2.01) | 1.50 (0.67–3.35) |
Number of participants could be smaller due to missing outcomes or population eligibility (screenings for hyperlipidemia, cervical cancer, breast cancer, and colon cancer).
Results are adjusted for propensity scores applied as inverse probability treatment weights, private/Medicaid insurance status, private/Medicaid HMO insurance status, poverty status, census region, MEPS panel year, and variables with standardized biases >0.25 (age, gender, marital status, percent of federal poverty line, language most often spoken at home, having a usual source of care, smoking status, distance to the nearest HC, household in a metropolitan statistical area and urban–rural continuum code) after propensity score adjustment.
Variables in propensity score estimation include the following: age, gender, marital status, race/ethnicity, education, percent of federal poverty line, language most often spoken at home, having a usual source of care, physical activity, smoking status, hypertension, heart disease, angina, myocardial infarction, stroke, emphysema, diabetes, arthritis, asthma, obesity, mental and physical health–related quality of life, depressed symptoms, number of functional limitations, household in a metropolitan statistical area, distance to the nearest health center, zip code–level poverty rate, zip code–level proportion of minority race, zip code–level proportion of Hispanics, urban–rural continuum code, MEPS survey year, and MEPS survey person weight variable.
***p < .001, **p < .01, *p < .05.
HMO, health management organization; IRR, incidence rate ratio; MEPS, Medical Expenditure Panel Survey; OR, odds ratio; SE, standard error.
We also compared health care utilization and receipt of preventive care between HC and non-HC patients among full-year uninsured and Medicaid subpopulations because of their high prevalence in HCs. We reestimated propensity scores and constructed regressions for subpopulations separately. In sensitivity analyses, using the same study population (adults ≥18 years, living ≤20 miles of an HC who had a clinic visit in the first panel year), we redefined HC patients to be participants with ≥1 HC visit versus those with no HC visits (non-HC patients). Propensity scores were estimated using SAS version 9.3 (SAS Institute), and regressions were performed using Stata version 12.1 (College Station, TX, USA). All statistical tests were conducted at the 5 percent two-sided significance level.
Results
Characteristics of Health Center and Non–Health Center Patients
About 3 percent of our study population (N = 33,137) had the majority of clinic visits at HCs (n = 976). Several differences were present between HC and non-HC patients (Table 1). Compared to non-HC patients, HC patients were younger (42.3 vs. 48.3 years), disproportionately minority race/ethnicity (67 percent vs. 27 percent), and more likely to be Spanish-speaking (28 percent vs. 5 percent) (all p < .001). Fewer HC patients had private insurance (28 percent vs. 73 percent, p < .001) or a usual source of care (79 percent vs. 84 percent, p = .009). HC patients had more current smokers (24 percent vs. 16 percent, p < .001) and a higher frequency of diabetes (14 percent vs. 10 percent, p = .004). HC patients lived closer to HCs than non-HC patients (2.8 vs. 6.0 miles, p < .001).
Health Care Utilization
In unadjusted analyses, HC patients had fewer office visits (4.67 vs. 7.57; incidence rate ratio [IRR], 0.62, p < .001) and more ER visits (0.31 vs. 0.21; IRR, 1.46, p = .003) than non-HC patients (Table 2). No differences were detected in hospital-based outpatient visits, prescriptions filled, or hospitalizations.
After adjustment, HC patients had fewer office visits (adjusted incidence rate ratio [aIRR], 0.63) and fewer hospitalizations (aIRR, 0.43) (both p < .001) relative to non-HC patients. ER visits were no longer greater among HC patients after adjustment. Rates of hospital-based outpatient visits and prescriptions filled were not different between HC and non-HC patients.
Receipt of Preventive Care
In unadjusted analyses, HC patients had lower odds of influenza vaccination (OR, 0.60, p < .001), hyperlipidemia screening (OR, 0.58, p = .03), and colon cancer screening (OR, 0.42, p < .001) compared to non-HC patients (Table 2). After adjustment, these differences were no longer significant. For breast cancer screening, unadjusted analysis suggested no difference (OR, 1.04, p = .89). However, after adjustment, HC patients had about a nearly three times higher odds of breast cancer screening than non-HC patients (aOR, 2.78, p = .001). No other differences in preventive care receipt were detected in unadjusted or adjusted results.
Uninsured Subpopulation
Uninsured HC patients (n = 359) were disproportionately racial/ethnic minorities (77 percent vs. 42 percent), less educated (10.6 vs. 12.3 years), and had higher rates of several chronic conditions than non-HC patients (n = 3,731) (e.g., myocardial infarction [6 percent vs. 2 percent] and diabetes [12 percent vs. 8 percent]) (all p < .001) (Table S2).
In unadjusted analyses, uninsured HC patients had fewer office visits (2.92 vs. 4.53; IRR, 0.65, p < .001) than uninsured non-HC patients, but no other differences in utilization. In adjusted analyses, uninsured HC patients had fewer office visits (aIRR, 0.70, p = .03), hospital-based outpatient visits (aIRR, 0.51, p = .02), and ER visits (aIRR, 0.59, p = .03) (Table 3), but no difference in hospitalizations.
Uninsured HC patients reported more preventive care across most measures in unadjusted and adjusted analyses (Table 3). In unadjusted analyses, uninsured HC patients had higher odds of dietary advice (OR, 1.72, p < .001), exercise advice (OR, 1.53, p = .006), cervical cancer screening (OR, 2.07, p = .005), and breast cancer screening (OR, 3.23, p = .005). In adjusted analyses, uninsured HC patients still had higher odds of dietary advice (aOR, 1.98, p = .002) and breast cancer screening (aOR, 4.53, p = .003). No differences in influenza vaccination, hypertension, hyperlipidemia, and colon cancer screening were present in unadjusted or adjusted analyses.
Medicaid Subpopulation
The Medicaid HC subpopulation was small, n = 174. HC patients were disproportionately racial/ethnic minorities (73 percent vs. 57 percent, p < .001) and had lower rates of several chronic diseases than non-HC patients (n = 1,774) (e.g., heart disease [0.8 percent vs. 4.6 percent], emphysema [0.6 percent vs. 4.1 percent], diabetes [12.7 percent vs. 13.4 percent]) (all p < .001) (Table S3).
In unadjusted analyses, HC patients with Medicaid insurance had fewer prescription drugs (17.5 vs. 26.2; IRR, 0.67, p = .04) compared to non-HC patients (Table 4). No differences were detected in office, hospital-based outpatient, or ER visits, or hospitalizations in unadjusted analyses. After adjustment, HC patients with Medicaid insurance had fewer hospital-based outpatient visits (aIRR, 0.45, p = .01), but more ER visits (aIRR, 1.49, p = .009) compared to non-HC patients. No differences in office and prescriptions filled were detected in adjusted analyses. Due to the reduced sample size, adjusted differences in hospitalization for the Medicaid subpopulation could not be estimated.
Table 4.
Health Care Utilization and Receipt of Preventive Care among Full-Year Medicaid Adult MEPS Participants (N = 1,948†)
| Health Center (n = 174†) | Non–Health Center (n = 1,774†) | Unadjusted IRR or OR (95% CI) | Adjusted IRR or OR (95% CI)‡,§ | |
|---|---|---|---|---|
| Health care utilization | ||||
| Office visits, mean (SE) | 6.62 (1.01) | 8.13 (0.51) | 0.81 (0.58–1.14) | 0.77 (0.56–1.05) |
| Hospital outpatient visits, mean (SE) | 0.46 (0.21) | 0.90 (0.13) | 0.51 (0.20–1.29) | 0.22 (0.13–0.40)*** |
| Prescription drugs, mean (SE) | 17.5 (3.17) | 26.2 (1.30) | 0.67 (0.46–0.98)* | 0.70 (0.50–0.98)* |
| Emergency room visits, mean (SE) | 0.53 (0.09) | 0.46 (0.03) | 1.15 (0.79–1.66) | 1.26 (0.97–1.65) |
| Hospitalizations, mean (SE) | 0.20 (0.05) | 0.26 (0.02) | 0.74 (0.45–1.23) | ¶ |
| Receipt of preventive care | ||||
| Diet advice, n (%) | 64 (32) | 693 (37) | 0.80 (0.51–1.26) | 0.90 (0.56–1.45) |
| Exercise advice, n (%) | 79 (41) | 801 (46) | 0.79 (0.51–1.23) | 0.98 (0.58–1.63) |
| Influenza vaccination, n (%) | 46 (28) | 525 (32) | 0.83 (0.50–1.40) | 0.80 (0.49–1.32) |
| Hypertension screening, n (%) | 159 (92) | 1,536 (89) | 1.33 (0.65–2.70) | 2.74 (1.33–5.64)** |
| Hyperlipidemia screening, n (%) | 53 (90) | 615 (95) | 0.53 (0.13–2.25) | 0.85 (0.20–3.55) |
| Cervical cancer screening, n (%) | 118 (94) | 1,052 (88) | 2.02 (0.84–4.85) | 1.03 (0.41–2.62) |
| Breast cancer screening, n (%) | 16 (85) | 195 (72) | 2.22 (0.44–11.1) | 1.08 (0.26–4.49) |
| Colon cancer screening, n (%) | 12 (48) | 152 (48) | 1.00 (0.32–3.10) | 2.18 (0.87–5.46) |
Number of participants could be smaller due to missing outcomes or population eligibility (screenings for hyperlipidemia, cervical cancer, breast cancer, and colon cancer).
Results are adjusted for propensity scores applied as inverse probability treatment weights, private/Medicaid insurance status, private/Medicaid HMO insurance status, poverty status, census region, MEPS panel year, and variables with standardized biases > 0.25 (race/ethnicity, percent of federal poverty line, language most often spoken at home, having a usual source of care, stroke, emphysema, urban–rural continuum code) after propensity score adjustment.
Variables in propensity score estimation include the following: age, gender, marital status, race/ethnicity, education, percent of federal poverty line, language most often spoken at home, having a usual source of care, physical activity, smoking status, hypertension, heart disease, angina, myocardial infarction, stroke, emphysema, diabetes, arthritis, asthma, obesity, mental and physical health–related quality of life, depressed symptoms, number of functional limitations, household in a metropolitan statistical area, distance to the nearest health center, zip code–level poverty rate, zip code–level proportion of minority race, zip code–level proportion of Hispanics, urban–rural continuum code, MEPS survey year, and MEPS survey person weight variable.
Model was unable to converge due to low prevalence of outcome.
***p < .001, **p < .01, *p < .05.
HMO, health management organization; IRR, incidence rate ratio; MEPS, Medical Expenditure Panel Survey; OR, odds ratio; SE, standard error.
There were no differences in receipt of preventive care between HC and non-HC patients with Medicaid insurance in unadjusted analyses. After adjustment, HC patients had higher odds of hypertension screening (aOR, 2.78, p = .03) compared to non-HC patients (Table 4).
Sensitivity Analyses
In sensitivity analyses, using the same study population (adults ≥18 years, living ≤20 miles of an HC who had a clinic visit in the first panel year), we redefined HC patients to be participants with ≥1 HC visit versus those with no HC visits (non-HC patients). In general, using the broader definition of HC patients, we found fewer differences in utilization between HC and non-HC patients. For the full population, HC patients had more ER visits (aIRR, 1.38, p = .04), and for the uninsured and Medicaid subpopulations, HC patients had fewer hospital-based outpatient visits. Preventive measures were reported at similar or better rates by HC patients compared to non-HC patients, across the full population and insurance subpopulations. For example, in the full population sensitivity analyses, HC patients had higher odds of breast cancer screening (aOR, 2.26, p = .009) compared to non-HC patients (Table 5) (see Tables S4 and S5 for uninsured and Medicaid subpopulation sensitivity analyses).
Table 5.
Sensitivity Analysis of Health Center Patient Definition: Health Care Utilization and Receipt of Preventive Care (N = 33,137†)
| HC (n = 1,413†,‡) | Non-HC (n = 31,724†) | Unadjusted IRR or OR (95% CI) | Adjusted IRR or OR (95% CI)§,¶ | |
|---|---|---|---|---|
| Health care utilization | ||||
| Office visits, mean (SE) | 7.59 (0.48) | 7.52 (0.09) | 1.01 (0.89–1.15) | 1.03 (0.87–1.21) |
| Hospital outpatient visits, mean (SE) | 0.69 (0.09) | 0.69 (0.03) | 0.99 (0.75–1.30) | 0.98 (0.65–1.50) |
| Prescription drugs, mean (SE) | 21.1 (1.21) | 16.4 (0.22) | 1.29 (1.15–1.44)*** | 1.08 (0.95–1.24) |
| Emergency room visits, mean (SE) | 0.36 (0.04) | 0.21 (0.004) | 1.71 (1.39–2.10)*** | 1.38 (1.02–1.86)* |
| Hospitalizations, mean (SE) | 0.16 (0.02) | 0.15 (0.004) | 1.12 (0.90–1.39) | 0.76 (0.55–1.05) |
| Receipt of preventive care | ||||
| Diet advice, n (%) | 640 (44) | 12,996 (41) | 1.14 (0.98–1.33) | 1.08 (0.83–1.40) |
| Exercise advice, n (%) | 682 (47) | 14,346 (45) | 1.08 (0.92–1.28) | 1.12 (0.85–1.46) |
| Influenza vaccination, n (%) | 499 (38) | 12,125 (41) | 0.87 (0.73–1.04) | 1.04 (0.81–1.34) |
| Hypertension screening, n (%) | 1,227 (91) | 26,927 (89) | 1.24 (0.94–1.62) | 1.70 (1.14–2.54)*** |
| Cholesterol screening, n (%) | 666 (96) | 18,135 (97) | 0.84 (0.55–1.27) | 1.52 (0.79–2.92) |
| Cervical cancer screening, n (%) | 747 (90) | 12,482 (89) | 1.10 (0.81–1.50) | 1.15 (0.69–1.90) |
| Breast cancer screening, n (%) | 255 (84) | 5,099 (83) | 1.09 (0.75–1.60) | 2.26 (1.23–4.16)*** |
| Colon cancer screening, n (%) | 191 (56) | 5,946 (70) | 0.55 (0.42–0.72)*** | 0.66 (0.43–1.00)* |
Number of participants could be smaller due to missing outcomes or population eligibility (screenings for cholesterol, cervical cancer, breast cancer, and colon cancer).
HC patients were those who received at least one clinic visit in a HC during the first panel year of MEPS.
Results are adjusted for propensity scores applied as inverse probability treatment weights, private/Medicaid insurance status, private/Medicaid HMO insurance status, poverty status, census region, MEPS panel year, and variables with standardized biases > 0.25 (age, Medicare insurance) after propensity score adjustment.
Variables in propensity score estimation include the following: age, gender, marital status, race/ethnicity, education, private/Medicare/Medicaid insurance, private/Medicaid HMO insurance, percent of federal poverty line, language most often spoken at home, having a usual source of care, physical activity, smoking status, hypertension, heart disease, angina, myocardial infarction, stroke, emphysema, diabetes, arthritis, asthma, obesity, mental and physical health–related quality of life, depressed symptoms, number of functional limitations, household in a metropolitan statistical area, distance to the nearest health center, zip code–level poverty rate, zip code–level proportion of minority race, zip code–level proportion of Hispanics, urban–rural continuum code, MEPS survey year, and MEPS survey person weight variable.
***p < .001, *p < .05.
HMO, health management organization; IRR, incidence rate ratio; MEPS, Medical Expenditure Panel Survey; OR, odds ratio; SE, standard error.
Discussion
Because of the expansion of the Section 330 HC program and impending growth of the insured population under the ACA, it is important to understand what additional value HCs provide to the health care system on a national scale beyond the care provided by other sites of primary care. We found that HC patients had fewer office visits and hospitalizations compared to other similar patients. In addition, HC patients had almost a three times higher odds of breast cancer screening than non-HC patients, and no other statistically significant differences in preventive care after adjustment. In our subpopulation analyses, the findings were particularly striking among uninsured patients, where in unadjusted and adjusted analyses, HC patients showed lower utilization but higher odds of receiving dietary advice and breast cancer screening.
Several explanations may underlie how HC patients have fewer office visits yet receive similar or more preventive care than patients at non-HCs. HC patients may have more barriers to physician specialists and thus have fewer visits to office-based providers. However, to compensate for the increased barriers, HCs may provide more care per visit compared to other sites of primary care, which would be aligned with the HC mission of providing comprehensive primary care. Also, HCs may have greater uncertainty about whether (and when) their patients will follow up, and so HCs may provide comprehensive care within each visit. Thus, HC patients may receive more preventive, acute, and chronic care per visit, which decreases the need for follow-up visits and hospitalizations.
Our findings of less utilization and similar or more preventive care among HC patients were particularly notable for the uninsured. HCs provide a unique primary health care access point for the uninsured, for whom the alternative may be no care. Thus, it is expected that the uninsured who receive care at HCs benefit from the increased access to primary and preventive care. In addition, the access to care provided by HCs may allow this vulnerable population to avoid increased severity of illnesses or complications that necessitate frequent outpatient follow-up visits and/or hospitalizations.
HC patients had strikingly higher reports of breast cancer screening compared to similar non-HC patients in this study. This finding was robust to a more broad definition of HC patients and also present in the uninsured. As HCs are closely tied to their communities, as part of their mission, they may be more aware of local resources for mammograms than other primary care sites. In addition, federally funded HCs may be more attuned than other primary care sites to governmental resources such as the National Breast and Cervical Cancer Early Detection Program (Centers for Disease Control and Prevention) for funding mammograms.
The differences between our unadjusted and adjusted results highlight the importance of adequate adjustment when comparing HC and non-HC populations. Health care payors and newly formed Accountable Care Organizations are increasingly evaluating the performance of primary care providers based on their ability to reduce unnecessary utilization and provide needed preventive care (Rosenthal et al. 2006; Addicott 2012). Without effective methods to adjust for differences in patient populations, HCs may be viewed as inadequately managing their patients compared to other primary care settings and, as a result, suffer negative financial penalties, similar to the unfavorable penalties that may be garnered by safety-net institutions by the Hospital Readmission Reduction Program (Joynt and Jha 2013). Our results suggest the importance of adequate adjustment for the patient and geographic factors that influence utilization across primary care settings.
Surprisingly, we found that among Medicaid beneficiaries, HC patients had higher rates of ER visits compared to non-HC patients. We believe these results should be interpreted with caution. These results run counter to two previous state-level Medicaid claims analyses, which found significantly fewer ambulatory care sensitive ER visits among HC patients (Falik et al. 2001, 2006). In addition, our results are based on a very small sample of Medicaid HC patients (n = 174), and while our adjustment addresses some variations in geography, it may not able to address the likely important state differences in Medicaid eligibility and coverage. Further study of the effects of HCs on health care utilization and preventive care among the Medicaid population is needed.
Our study has several limitations. Our results are not generalizable to all HCs or to HCs that do not receive federal funding (such as HC Look-Alikes). In addition, our results are not generalizable to all HC patients, as participants had to live within a household for 2 years to be included in MEPS. In addition, we cannot ensure that all non-HC patients received primary care, as there is no clear measure of primary care in the MEPS. However, it is likely that most non-HC patients received some primary care, as the non-HC population reported a higher rate of having a usual source of care (84 percent vs. 78 percent) compared to the HC population. Our study also only evaluated differences in health care utilization 1 year after having outpatient care. It is possible that a longer study may find that the lower utilization among HC patients may not persist or lead to negative long-term clinical outcomes. Also, self-reported health care utilization and receipt of preventive care are subject to recall bias, which may not be adequately addressed by our adjustment for measured covariates. Finally, while we believe that our method of HC assignment via address matching is an improvement over previous analyses, the method still has limitations, as HC sites may be colocated with non-HC providers.
Overall, our national study found that patients who receive the majority of their care at HCs have less subsequent utilization of outpatient and inpatient care, while having higher odds of breast cancer screening. The dynamic of reduced utilization and improved preventive care is particularly prominent among uninsured patients. These intriguing findings suggest that federally funded HCs add value to the health care system by serving socially and medically disadvantaged patients with care that results in lower system-wide utilization and maintained or improved preventive care services.
Acknowledgments
Joint Acknowledgment/Disclosure Statement: Dr. Laiteerapong had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the analysis. Dr. Laiteerapong led the study concept and design, analysis and interpretation of data, and preparation of the manuscript. Dr. Kirby, Ms. Gao, Dr. Yu, Dr. Lee, and Dr. Huang participated in the study concept and design, analysis and interpretation of data, and preparation of manuscript. Ms. Gao participated in the analysis and interpretation of data and preparation of the manuscript. Drs. Sharma and Ngo-Metzger participated in the study concept and design, analysis and interpretation of data, and preparation of the manuscript. Mr. Nocon and Dr. Chin participated in the study concept and design and preparation of the manuscript. Ms. Nathan participated in the analysis and interpretation of data and the preparation of manuscript. Funding for this project was sponsored by the Health Resources and Services Administration’s Bureau of Primary Health Care. The sponsor (BPHC) was involved in the design and conduct of the study, analysis and interpretation of the data, and review and approval of the manuscript. The sponsor was not involved in the collection or management of data. Dr. Laiteerapong is supported by NIDDK K23 DK097283 and a John A. Hartford Foundation Center of Excellence Award (American Federation for Aging Research). Dr. Chin is supported by NIDDK K24 DK071933. Dr. Laiteerapong, Ms. Gao, Mr. Nocon, Ms. Nathan, Dr. Chin, and Dr. Huang are members of the NIDDK Chicago Center for Diabetes Translation Research at the University of Chicago (P30 DK092949). This manuscript was presented in abstract form to the Society of General Internal Medicine (SGIM) 36th Annual Meeting, Denver, Colorado, April 2013. The authors thank Elizabeth A. Stuart, PhD, Department of Mental Health and Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, for providing consultation on the use of propensity score methods with survey data. Dr. Stuart received compensation for her work.
Disclosures: None.
Disclaimers: None.
Supporting Information
Additional supporting information may be found in the online version of this article:
Appendix SA1: Author Matrix.
Data S1: Calculation of Distance between Households and Health Centers, Accounting for Missing Values Using Hot-Deck Imputation, Evaluation of Variable Balance of Propensity Score Estimation.
Table S1: Standardized Bias Values of Propensity Score Variables.
Table S2: Full-Year Uninsured Adult Participants Living within Twenty Miles of a Health Center Receiving Outpatient Care in the First Panel Year.
Table S3: Full-Year Medicaid-Insured Adult Participants Living within Twenty Miles of a Health Center Receiving Outpatient Care in the First Panel Year.
Table S4: Sensitivity Analysis of Health Center Patient Definition among Full-Year Uninsured Adult MEPS Participants: Health Care Utilization and Receipt of Preventive Care.
Table S5: Sensitivity Analysis of Health Center Patient Definition among Full-Year Medicaid Adult MEPS Participants: Health Care Utilization and Receipt of Preventive Care.
References
- Addicott R. What Accountable Care Organizations Will Mean for Physicians. British Medical Journal. 2012;345:e6461. doi: 10.1136/bmj.e6461. [DOI] [PubMed] [Google Scholar]
- Centers for Disease Control and Prevention. “National Breast and Cervical Cancer Early Detection Program (NBCCEDP)” [accessed on May 7, 2013]. Available at http://www.cdc.gov/cancer/nbccedp/
- D’Agostino RB., Jr Propensity Score Methods for Bias Reduction in the Comparison of a Treatment to a Naon-Randomized Control Group. Statistics in Medicine. 1998;17(19):2265–81. doi: 10.1002/(sici)1097-0258(19981015)17:19<2265::aid-sim918>3.0.co;2-b. [DOI] [PubMed] [Google Scholar]
- Dugoff EH, Schuler M. Stuart EA. Generalizing Observational Study Results: Applying Propensity Score Methods to Complex Surveys. Health Services Research. 2013;49:284–303. doi: 10.1111/1475-6773.12090. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Epstein AJ. The Role of Public Clinics in Preventable Hospitalizations among Vulnerable Populations. Health Services Research. 2001;36(2):405–20. [PMC free article] [PubMed] [Google Scholar]
- Falik M, Needleman J, Wells BL. Korb J. Ambulatory Care Sensitive Hospitalizations and Emergency Visits: Experiences of Medicaid Patients Using Federally Qualified Health Centers. Medical Care. 2001;39(6):551–61. doi: 10.1097/00005650-200106000-00004. [DOI] [PubMed] [Google Scholar]
- Falik M, Needleman J, Herbert R, Wells B, Politzer R. Benedict MB. Comparative Effectiveness of Health Centers as Regular Source of Care: Application of Sentinel ACSC Events as Performance Measures. Journal of Ambulatory Care Management. 2006;29(1):24–35. doi: 10.1097/00004479-200601000-00004. [DOI] [PubMed] [Google Scholar]
- Forrest CB. Whelan EM. Primary Care Safety-Net Delivery Sites in the United States: A Comparison of Community Health Centers, Hospital Outpatient Departments, and Physicians’ Offices. Journal of the American Medical Association. 2000;284(16):2077–83. doi: 10.1001/jama.284.16.2077. [DOI] [PubMed] [Google Scholar]
- Gilbody S, Richards D, Brealey S. Hewitt C. Screening for Depression in Medical Settings with the Patient Health Questionnaire (PHQ): A Diagnostic Meta-Analysis. Journal of General Internal Medicine. 2007;22(11):1596–602. doi: 10.1007/s11606-007-0333-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Goldman LE, Chu PW, Tran H, Romano MJ. Stafford RS. Federally Qualified Health Centers and Private Practice Performance on Ambulatory Care Measures. American Journal of Preventive Medicine. 2012;43(2):142–9. doi: 10.1016/j.amepre.2012.02.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gresenz CR, Rogowski J. Escarce JJ. Health Care Markets, the Safety Net, and Utilization of Care among the Uninsured. Health Services Research. 2007;42(1 Pt 1):239–64. doi: 10.1111/j.1475-6773.2006.00602.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gurewich D, Tyo KR, Zhu J. Shepard DS. Comparative Performance of Community Health Centers and Other Usual Sources of Primary Care. Journal of Ambulatory Care Management. 2011;34(4):380–90. doi: 10.1097/JAC.0b013e31822cbc59. [DOI] [PubMed] [Google Scholar]
- Gurewich D, Capitman J, Sirkin J. Traje D. Achieving Excellence in Community Health Centers: Implications for Health Reform. Journal of Health Care for the Poor and Underserved. 2012;23(1):446–59. doi: 10.1353/hpu.2012.0008. [DOI] [PubMed] [Google Scholar]
- Hadley J. Cunningham P. Availability of Safety Net Providers and Access to Care of Uninsured Persons. Health Services Research. 2004;39(5):1527–46. doi: 10.1111/j.1475-6773.2004.00302.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Harder VS, Stuart EA. Anthony JC. Propensity Score Techniques and the Assessment of Measured Covariate Balance to Test Causal Associations in Psychological Research. Psychological Methods. 2010;15(3):234–49. doi: 10.1037/a0019623. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hicks LS, O’Malley AJ, Lieu TA, Keegan T, Cook NL, McNeil BJ, Landon BE. Guadagnoli E. The Quality of Chronic Disease Care in U.S. Community Health Centers. Health Affairs (Millwood) 2006;25(6):1712–23. doi: 10.1377/hlthaff.25.6.1712. [DOI] [PubMed] [Google Scholar]
- Joynt KE. Jha AK. A Path Forward on Medicare Readmissions. New England Journal of Medicine. 2013;368(13):1175–7. doi: 10.1056/NEJMp1300122. [DOI] [PubMed] [Google Scholar]
- Moyer VA. Screening for Cervical Cancer: U.S. Preventive Services Task Force Recommendation Statement. Annals of Internal Medicine. 2012;156(12):880–91. doi: 10.7326/0003-4819-156-12-201206190-00424. [DOI] [PubMed] [Google Scholar]
- Porterfield DS. Kinsinger L. Quality of Care for Uninsured Patients with Diabetes in a Rural Area. Diabetes Care. 2002;25(2):319–23. doi: 10.2337/diacare.25.2.319. [DOI] [PubMed] [Google Scholar]
- Probst JC, Laditka JN. Laditka SB. Association between Community Health Center and Rural Health Clinic Presence and County-Level Hospitalization Rates for Ambulatory Care Sensitive Conditions: An Analysis across Eight US States. BMC Health Services Research. 2009;9:134. doi: 10.1186/1472-6963-9-134. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Regan J, Schempf AH, Yoon J. Politzer RM. The Role of Federally Funded Health Centers in Serving the Rural Population. Journal of Rural Health. 2003;19(2):117–24. doi: 10.1111/j.1748-0361.2003.tb00552.x. discussion 15–6. [DOI] [PubMed] [Google Scholar]
- Reilly M. Pepe M. The Relationship between Hot-Deck Multiple Imputation and Weighted Likelihood. Statistics in Medicine. 1997;16(1–3):5–19. doi: 10.1002/(sici)1097-0258(19970115)16:1<5::aid-sim469>3.0.co;2-8. [DOI] [PubMed] [Google Scholar]
- Richard P, Ku L, Dor A, Tan E, Shin P. Rosenbaum S. Cost Savings Associated with the Use of Community Health Centers. Journal of Ambulatory Care Management. 2012;35(1):50–9. doi: 10.1097/JAC.0b013e31823d27b6. [DOI] [PubMed] [Google Scholar]
- Rosenbaum PL. Rubin DB. The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika. 1983;70(1):41–55. [Google Scholar]
- Rosenthal MB, Landon BE, Normand SL, Frank RG. Epstein AM. Pay for Performance in Commercial HMOs. New England Journal of Medicine. 2006;355(18):1895–902. doi: 10.1056/NEJMsa063682. [DOI] [PubMed] [Google Scholar]
- Rothkopf J, Brookler K, Wadhwa S. Sajovetz M. Medicaid Patients Seen at Federally Qualified Health Centers Use Hospital Services Less Than Those Seen by Private Providers. Health Affairs (Millwood) 2011;30(7):1335–42. doi: 10.1377/hlthaff.2011.0066. [DOI] [PubMed] [Google Scholar]
- Rubin DB. Estimating Causal Effects from Large Data Sets Using Propensity Scores. Annals of Internal Medicine. 1997;127(8 Pt 2):757–63. doi: 10.7326/0003-4819-127-8_part_2-199710151-00064. [DOI] [PubMed] [Google Scholar]
- Rust G, Baltrus P, Ye J, Daniels E, Quarshie A, Boumbulian P. Strothers H. Presence of a Community Health Center and Uninsured Emergency Department Visit Rates in Rural Counties. Journal of Rural Health. 2009;25(1):8–16. doi: 10.1111/j.1748-0361.2009.00193.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Scherer TM. Lewis LM. Follow-Up to a Federally Qualified Health Center and Subsequent Emergency Department Utilization. Academic Emergency Medicine. 2010;17(1):55–62. doi: 10.1111/j.1553-2712.2009.00621.x. [DOI] [PubMed] [Google Scholar]
- Shi L. Stevens GD. The Role of Community Health Centers in Delivering Primary Care to the Underserved: Experiences of the Uninsured and Medicaid Insured. Journal of Ambulatory Care Management. 2007;30(2):159–70. doi: 10.1097/01.JAC.0000264606.50123.6d. [DOI] [PubMed] [Google Scholar]
- Shi L, Lebrun LA, Tsai J. Zhu J. Characteristics of Ambulatory Care Patients and Services: A Comparison of Community Health Centers and Physicians’ Offices. Journal of Health Care for the Poor and Underserved. 2010;21(4):1169–83. doi: 10.1353/hpu.2010.0928. [DOI] [PubMed] [Google Scholar]
- Smith-Campbell B. Emergency Department and Community Health Center Visits and Costs in an Uninsured Population. Journal of Nursing Scholarship. 2005;37(1):80–6. doi: 10.1111/j.1547-5069.2005.00011.x. [DOI] [PubMed] [Google Scholar]
- Starfield B, Powe NR, Weiner JR, Stuart M, Steinwachs D, Scholle SH. Gerstenberger A. Costs vs Quality in Different Types of Primary Care Settings. Journal of the American Medical Association. 1994;272(24):1903–8. [PubMed] [Google Scholar]
- Stuart EA. Matching Methods for Causal Inference: A Review and a Look Forward. Statistical Science. 2010;25(1):1–21. doi: 10.1214/09-STS313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ware J, Jr, Kosinski M. Keller SD. A 12-Item Short-Form Health Survey: Construction of Scales and Preliminary Tests of Reliability and Validity. Medical Care. 1996;34(3):220–33. doi: 10.1097/00005650-199603000-00003. [DOI] [PubMed] [Google Scholar]
- Zanutto EL. A Comparison of Propensity Score and Linear Regression Analysis of Complex Survey Data. Journal of Data Science. 2006;4(1):67–91. [Google Scholar]
- Zanutto E, Lu B. Hornik R. Using Propensity Score Subclassification for Multiple Treatment Doses to Evaluate a National Antidrug Media Campaign. Journal of Educational and Behavioral Statistics. 2005;30(1):59–73. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Appendix SA1: Author Matrix.
Data S1: Calculation of Distance between Households and Health Centers, Accounting for Missing Values Using Hot-Deck Imputation, Evaluation of Variable Balance of Propensity Score Estimation.
Table S1: Standardized Bias Values of Propensity Score Variables.
Table S2: Full-Year Uninsured Adult Participants Living within Twenty Miles of a Health Center Receiving Outpatient Care in the First Panel Year.
Table S3: Full-Year Medicaid-Insured Adult Participants Living within Twenty Miles of a Health Center Receiving Outpatient Care in the First Panel Year.
Table S4: Sensitivity Analysis of Health Center Patient Definition among Full-Year Uninsured Adult MEPS Participants: Health Care Utilization and Receipt of Preventive Care.
Table S5: Sensitivity Analysis of Health Center Patient Definition among Full-Year Medicaid Adult MEPS Participants: Health Care Utilization and Receipt of Preventive Care.