The MEPS utilized a multistage, probability clustering sample design that enabled comprehensive examination of the US population. The sampling weight, stratification, clustering, multiple stages of selection, and disproportionate sampling from the MEPS were taken into account in the analyses so that the findings reported represented the entire US population. Sampling weight took into account the differential probability of sample selection and adjusted for nonresponses and missing data. Descriptive statistics were conducted to examine SAT, PCS, MCS, GH scores, EXP, and different service categories of medical expenditures across clinical conditions. We also performed descriptive statistics on all potential confounders and flagged those that had significant associations with patient satisfaction to be included as covariates in subsequent regression analyses.

To investigate the contribution of PCS, MCS, GH, and EXP on patient satisfaction, we conducted logistic regressions with adjustment of covariates and reported the odds ratio with associated 95% CI. We recoded the PROs, satisfaction, and expenditure variables into low/high SAT, PCS, MCS, GH, and EXP prior to running the logistic regressions. All statistical tests were two sided, were set at an alpha level of .05, and were conducted using SAS 9.3. Institutional review board (IRB) and/or ethics committee approval was not required as the MEPS data are freely available to the general public online.

The entire study sample consisted of adult participants (N=10,157), which represented 233.26 million adults (aged ≥ 18 years) in the United States. The mean age was 47 years (SE 0.28, range 18-85). Approximately 52% (121 million/233.26 million) were female and the majority were white (81%, 188 million/233.26 million) and black (12%, 27 million/233.26 million). About 15% (34 million/233.26 million) were Hispanic, approximately 34% (79 million/233.26 million) were unemployed, and nearly 12% (27 million/233.26 million) spoke non-English at home (Table 1).

This nationally representative adult population had a number of health conditions. About 32% (74.97 million/233.09 million) had high blood pressure and 5% (12.31 million/233.11 million) had coronary heart disease. Approximately 4% (8.52 million/233.20 million) had previously experienced a stroke. Almost 30% (70.59 million/233.07 million) had high cholesterol and 24% (56.34 million/233.18 million) had arthritis. There were 10% (23.48 million/233.20 million) diagnosed with cancer.

The sample population visited a variety of providers; however, most visited a general/family practice physician (71%, 60.69 million/86.00 million) and internal medicine physicians (20%, 17.10 million/86.00 million) within the last 12 months. About 68% (101.57 million/150.09 million) stated that their providers always showed respect for their treatment, whereas 3% (3.89 million/150.09 million) stated that their providers never showed respect. Almost 96% (161.35 million/168.25 million) indicated that their providers explained treatment options but 4% (6.90 million/168.25 million) did not. Approximately 67% (156.18 million/233.26 million) had private insurance, 18% (41.31 million/233.26 million) had public health insurance, whereas over 15% (35.77 million/233.26 million) were uninsured. A detailed breakdown of the number of responses for each variable is not provided here but can be obtained by contacting the authors.

The 2010 annual total health care expenditure in the US was over US $1.11 trillion and the average total health care expenditure per adult was US $4752.39 (Table 2). The total medical expenditures for emergency visits, hospital stay, dental care, and prescription drugs were US $7.17 billion, US $41.21 billion, US $62.87 billion, and US $252.45 billion, respectively. The US medical expenditures, by major service categories, by disease conditions are displayed in Table 2. Table 3 presents the descriptive statistics for medical expenditure, outcomes, and patient satisfaction.

The mean score for PCS, MCS, and GH was 49.28 (SE 0.16), 50.98 (SE 0.14), and 3.55 (SE 0.02), respectively. These scores were significantly correlated with each other (PCS with MCS, ρ=.16, P<.001; PCS with GH, ρ=.96, P<.001; MCS with GH, ρ=.62, P<.001).

As measured by CAHPS’s global rating, the mean patient satisfaction score across the national adult sample was 8.30 (SE 0.03). The following were some of the variables that were significantly associated with patient satisfaction: age (ρ=.25, P<.001), currently smoking (χ² ₁=29.7, P<.001), provider showed respect for treatment (χ² ₃=75.5, P<.001), and provider explained option to person (χ² ₁=21.0, P<.001). The complete list of all of the variables that were found to be associated with patient satisfaction is provided in Multimedia Appendix 2.

Total health care expenditure was negatively related to PROs; individuals who had lower physical function (ρ=−.43, P<.001), lower mental function (ρ=−.11, P<.001), and lower general health (ρ=−.51, P<.001) had higher total health care expenditures. After adjusting for covariates, we found that PROs and total health care expenditure were highly related to patient satisfaction (Table 4). Higher physical function, higher mental function, higher general health status, and higher total health care expenditure were associated with higher patient satisfaction. The odds of those who had high GH being satisfied were 6 times greater than those who had low GH (adjusted OR 5.98, 95% CI 2.95-12.12). High GH was defined as those who had excellent GH and low GH was defined as those who had poor or fair GH. There was more than a 2-fold difference in patient satisfaction between those who had high and low PCS (adjusted OR 2.54, 95% CI 1.36-4.72) and a 2-fold difference between those who had high and low MCS (adjusted OR 1.95, 95% CI 1.20-3.15). Those who had high EXP being satisfied were 3 times greater than those who had low EXP (adjusted OR 3.20, 95% CI 1.47-6.98). High PCS, MCS, and EXP were defined by those respondents who had ≥75 percentile of the scores, whereas low PCS, MCS, and EXP were defined by those with <25 percentile scores.

The purpose of this study was to investigate the contribution of physical health, mental health, general health, and total health care expenditures to patient satisfaction. We found that higher scores of physical health, mental health, and general health were related to higher scores of patient satisfaction. Consistent with previous research, we found that some factors such as age and provider interactions could impact patient satisfaction. Additionally, total health care expenditure was associated with patient satisfaction. Taken together, findings from this study can help providers, payers, policy makers, and the general public better understand the relationship between PROs, health care expenditures, and satisfaction.

General health demonstrated the strongest relationship with patient satisfaction. After accounting for demographics and various factors, we see that greater patient satisfaction is directly related to greater physical and mental health, although the relationship is stronger with physical health. The argument that the relation between higher patient satisfaction and better health is an artifact of the tendency of healthy patients to be satisfied is not new [21]. This stresses the importance of controlling for prior health status as we did in this study and with the inclusion of prior clinical conditions into the model. Previous research found that a functional measure of health predicted greater satisfaction levels immediately following the medical treatment [9]. Differing research results regarding the impact of patient satisfaction on health have been partially attributed to differences in how risk or health have been assessed as well as the time frame as to when the satisfaction scores were gathered [22]. It may be that individuals who are in poor health have to visit a physician more often, increasing the likelihood of a bad or unsatisfactory visit. Prior experience with health care, along with age and mental status variables had been shown to impact patient expectations [23]. Our study emphasizes the importance of controlling for prior health status, prior health care experience, nonmodifiable characteristics, and patient expectations when assessing the effect of patient satisfaction.

Mental health often goes unnoticed or it is dismissed as someone having a bad day. Both identifying mental health issues and addressing problems require training that goes beyond that of a general physician. As a result, a patient or a provider often needs to raise a concern about mental health. A stigma still exists about discussing and treating mental health issues, which may lead patients to be less willing to discuss mental health conditions. Seeing a mental health specialist adds an additional cost for patients that they may want to avoid. Finally, patients might not even notice that they have mental health issues. These may be reasons why mental health is a weaker predictor of patient satisfaction than physical health.

A surprising finding was that total health care expenditures also had a significant relation to patient satisfaction. In a recent review of health care quality, only one third of the 61 studies reviewed found a positive association between higher spending and better health care quality [10], but did not address patient satisfaction directly as a measure of quality. While our findings add to the body of research that suggests an association between spending and patient satisfaction [11], it contradicts other research that states there is no association [9,24]. It is possible that initiatives that improve PROs and control cost at the same time contribute to patient satisfaction. Instead of raising health care expenditures, health policy makers and providers should attend to modifiable characteristics that are within their control such as providers showing respect and providers explaining treatment options to enhance PROs and thus, patient satisfaction.

We were limited by the measurement tools in the MEPS, thus limiting the data we could use to measure physical, mental, and general health. Newer instruments have been developed using advanced techniques and these instruments may have better psychometric properties than instruments included in MEPS. The MEPS was administered using computer-assisted personal interviewing technology and the results obtained from this mode of survey administration may be different from those obtained from other modes. Care needs to be taken when doing cross-comparisons of results from different modes in different studies. Furthermore, the definition of health care expenditures reflects the CMS’s definition but might not represent a complete assessment of medical expenditures. We were not able to look at diseases that were not included in MEPS. Even though we had adjusted for nonresponses and missing data in the sample, such adjustment could only be made to a certain degree. The adjustment might not be adequate in the event that nonrespondents differed from respondents. It is well-known that there could be a larger share of nonrespondents having severe illness, impairment, and dealing with poor social economic conditions on a daily basis. Additionally, the majority of the population is white, but within the past few years, the United States has seen a growth of racial and ethnic diversity. Therefore, our findings might not represent underrepresented minorities at this moment in time.

Finally, the concept of patient satisfaction lacks a clear connotation and definition in the literature. It has sometimes been considered as a component of consumer marketing [6] that measures important customer service qualities within the realm of health care. These qualities include effective provider communication, support from physicians, waiting time, fulfillment of patient requests, staff integrity, and shared decision making, and represent multiple dimensions of patient satisfaction. Together, these dimensions of patient satisfaction constitute a patient’s overall health care experience. It is this overall patient health care experience that this study used to define patient satisfaction. It is likely that we were not accounting for everything that comprises patient satisfaction. In the future it would be beneficial to collect different aspects of patient satisfaction data and examine them under a different lens.

Future research should consider other modifiable characteristics that may influence patient satisfaction so that changes can be enacted. Often, underrepresented minorities have different experiences with health care than European Americans. Therefore, investigating their perspectives may illuminate group differences. Finally, future research should apply business intelligence or big data analytics to investigate whether there is a differential effect of PROs and medical expenditures on patient satisfaction across different medical areas.

We found that PROs and total health care expenditure had a strong relationship with patient satisfaction. As more emphasis is placed on health care value and quality, this area of research will become increasingly needed. Critical questions should be asked about what we value in health care and whether we can find a balance between patient satisfaction, outcomes, and expenditures. These questions need to be asked to policy makers, physicians, patients, insurers, and the general public.