Academic Editor: abha tewari
Affilation: The george Institute for Global Health, India
Using proprietary data of patient records from four medical clinics in the Mississippi Delta, this research utilizes a natural experiment design to explore if the patient centered medical home (PCMH) has a positive effect on chronic disease maintenance for low SES, majority African-American patients in a rural and medically underserved region. The patients are divided into two cohorts, those attending PCMH clinics (level 2) and those attending non-PCMH clinics. Each cohort is comprised of similar demographic, socioeconomic, and health (large proportion of diabetics) characteristics. HbA1c scores of the cohorts are compared at two time periods, baseline and six-month follow-up. PCMH patients report more uncontrolled diabetes at baseline but the trend reverses at follow-up, providing evidence that the PCMH model of primary care produces positive health outcomes for patients with diabetes in the sample area.
Under the current model of primary care delivery in the United States, nearly three out of four American adults report difficulty getting an appointment, health care advice by phone, or off-hours care without going to an emergency room. In order to meet this growing need for a more efficient and effective
The current evidence suggests that the PCMH model is working, but there remains an important gap in the literature regarding PCMH effectiveness in rural communities. Rural communities exhibit higher rates of chronic illness and mortality than urban communities,; and racial and ethnic minorities in rural communities exhibit higher rates of disease and death than white rural Americans, making them doubly disadvantaged. Our research fills this place- and race-based gap in the literature by examining HbA1c improvement over time for patients with diabetes between PCMH and
Why PCMH Matters in Mississippi and Rural America More Broadly
The Mississippi Delta is a prime example of a rural region experiencing a dearth of health care availability, as the majority of Delta counties are classified as health professional shortage areas. The shortage of physicians is particularly damaging because Delta counties exhibit high rates of hospitalizations due to diabetes complications. Additionally, the targeted service area has a majority African-American population, who experiences significantly higher incidence rates of type 2 diabetes and premature death than non-Hispanic whites, in large part a result of disparities in access and quality of care.; The region is also among the poorest in the nation, creating another significant layer of health disadvantage. Overall, given the Delta’s complex combination of medical, social, and economic disadvantages, it represents a community in dire need of health intervention.
We believe that if the PCMH model can be effective in the Mississippi Delta, among a population of medically underserved and majority minority and
We examine the effectiveness of the PCMH model of primary care on patient outcomes as a natural experiment. Patient records come from six Mississippi Delta clinics divided into two cohorts, PCMH (two clinics, both classified as level 2 PCMH clinics) and
The data include de-identified person-level information on clinic classification (PCMH or not), patient age (in years), race (African American or not), and patient clinical data associated with diabetes (HbA1c, systolic/diastolic blood pressure, LDL). Sex is not included in the analysis due to the unavailability for a large amount of patients. Some data (from both PCMH and
Regarding the health status of each cohort, cohort 1 (PCMH) has 87% of its patients exhibiting a mean T1HbA1c level of 6.5 or higher, and cohort 2 (
Table 1 provides background descriptive statistics for the PCMH and non-PCMH cohorts. The non-PCMH cohort includes four clinics, and therefore had a larger patient load with all of the qualifying data (primarily based on more than one clinic visit) (N = 1,040) than the two-clinic PCMH cohort (N = 143). The demographic makeup of the two cohort populations is similar, as both serve primarily African American patients (PCMH: 85.3%; non-PCMH: 74.9%). On average, PCMH patients were 10 years younger than non-PCMH patients (mean of 51 years old versus mean of 61 years old, respectively). Prior research suggests that this age difference may be insignificant from a clinical standpoint considering that patients above the age of 45 are considered the highest risk of developing diabetes. These patients are quite unhealthy regardless of age, as indicated by a T1 cohort-level percentage of diabetics (6.5 HbA1c) of 87% (PCMH) and 70% (non-PCMH). However, these proportions changed at the 6-month follow-up: 70% of PCMH patients exhibited diabetic HbA1c levels while 77% of non-PCMH patients exhibited diabetic HbA1c levels. Both patient cohorts were similar in terms of systolic blood pressure levels, diastolic blood pressure levels, and LDL readings (Table1).
Logistic regression models predicting the odds of being diabetic controlling for clinic setting, age, race, systolic/diastolic blood pressure, and LDL levels were performed at T1 (Table 2) and 6-month follow-up (Table 3). Table 2 shows that the only associated variable with being diabetic at T1 (p<.05) is age (OR = 0.96; 95% CI = 0.93 – 0.981; p <0.001), controlling for clinic setting, race, and T1 clinical scores. In this analysis clinic setting is approaching significance but does not meet the .05 threshold.
Table 3 shows the odds of being diabetic at 6-month follow-up, controlling for clinic setting, age, race, and T1 clinical scores. In this model, a key change in association occurs. Age continues to be significantly associated with likelihood of being diabetic, but unlike the previous model, clinical setting is now significantly associated at the .05 level as well (OR = 0.49; 95% CI = 0.27 to 0.88; p =0.017). A comparison of the odds ratios of clinic across the two models reveal that PCMH patients have higher average HbA1c scores at baseline and significantly lower average HbA1c scores at 6-month follow-up, controlling for age, race, and T1 blood pressure and cholesterol scores. In other words, the cohort of PCMH patients was unhealthier than that of non-PCMH patients at T1 but is healthier than non-PCMH patients at 6-month follow-up.
Additional analyses are conducted to confirm the original results. Both logistic regression models are fitted to include only clinic status and age. The parameter estimates for the newly fitted models are shown in Tables 4 and 5, and both clinic setting and age remain significantly associated with diabetic HbA1c levels at T1 and 6-month follow-up. Furthermore, the interaction between clinic setting and age is explored to further parse out the role of age in predicting diabetic status, and no significant interaction effect is found at either T1 (p > 0.42) or 6-month follow-up (p > 0.78). This provides further evidence that improvement in HbA1c is due to clinic setting rather than age or any other control variable. The positive result for diabetics in PCMH clinics is robust to several different statistical models.
This research demonstrates that the
Secondly, there is a large body of research suggesting that access to health care and health disparities are highly dependent on place. Whether “place” is operationalized as rural-urban, south-non-south, state, county, or regional differences, it is clear that context matters. Despite the built-in place-based challenges this target population faces, the PCMH model dramatically improved effective maintenance of diabetes. If the PCMH model can be effective in this place, we argue that it can be effective in other places with high levels of need. These findings are particularly generalizable to other parts of rural America. Rural Americans exhibit higher rates of morbidity and mortality, suffer from chronic illnesses at higher rates than urban Americans, and have more difficulty in accessing care, particularly the case for the chronically ill.; Many rural communities are looking for methods to enhance patient outcomes with improved delivery of care, and these findings address a critical gap in the literature by showing the effectiveness of the PCMH model in the Mississippi Delta community.
Third, it should come as no surprise that the PCMH model is effective. By definition, the model engages the patients in their own care, provides a more complex network of care, and increased number of resources through which to seek care, maintain health, and improve well-being. Considering that there are clinic-level factors that cannot be accounted for in statistical models, such as clinic leadership, management, and qualitative differences in quality of care, there is one key component that can be measured, adherence to follow-up appointments. The patient-centered approach in our “experimental” clinics is most noticeably evident in adherence to follow-up appointments. Adherence is vital to maintaining chronic disease, yet only 25% of diabetic patients nationwide adhere to six-month follow-up appointments. Diabetic patients in this study’s
Finally, the results of this work have contributed to existing gaps in the current PCMH literature. Prior work has consistently found that the PCMH model is more financially advantageous than the traditional primary care model; it has also been shown to be beneficial for patient perception of the care they receive, but clinical outcomes as it pertains to PCMH effectiveness has not been extensively studied thus far. In addition, there is a lack of clinical evidence that has been analyzed for rural Americans in general, much less specific sub-populations of this group. If indeed the CDC and AAFP are correct in their assessments that the nation is facing a mortality crisis rooted in racial and socioeconomic disparities, then this work has uncovered valuable information about the ways in which disadvantaged populations can potentially improve their conditions.
The results of this research provide evidence of the effectiveness of the PCMH model for diabetes maintenance in a region in desperate need of improved health care availability and delivery. The residents of the Mississippi Delta suffer from levels of chronic disease higher than the national average, while also residing in a health professional shortage area that exhibits high rates of poverty and inequality. This description is applicable to much of rural America, particularly in the Southern states. This work extends what is currently known about
Another limitation is the use of the 6.5 HbA1c
We are aware of the limitation of the small sample size of our study, but we argue that the results shown are significant enough to warrant discussion. There is a need for further analysis of populations like that of the Mississippi Delta in order to determine the effectiveness of the PCMH model in other rural areas that may exhibit somewhat different characteristics from ours. We believe this evaluation can be a crucial first step in opening the door to a broader discussion on how to decrease the burden of mortality and income-based disparities.