Dutch healthcare reform: did it result in better patient experiences in hospitals? a comparison of the consumer quality index over time

Background In 2006, the Dutch hospital market was reformed to create a more efficient delivery system through managed competition. To allow competition on quality, patient experiences were measured using the Consumer Quality index (CQI). We study whether public reporting and competition had an effect on the CQI between 2006 and 2009. Methods We analyzed 8,311 respondents covering 31 hospitals in 2006, 22,333 respondents covering 78 hospitals in 2007 and 24,246 respondents covering 94 hospitals in 2009. We describe CQI trends over the period 2006-2009. In addition we compare hospitals that varied in the level of competition they faced and hospitals that were forced to publish CQI results publicly and those that were not. We corrected for observable covariates between hospital respondents using a multi level linear regression. We used the Herfindahl Hirschman Index to indicate the level of competition. Results Between 2006 and 2009 hospitals showed a CQI improvement of 0.034 (p < 0.05) to 0.060 (p < 0.01) points on a scale between one and four. Hospitals that were forced to publish their scores showed a further improvement of 0.027 (p < 0.01) to 0.030 (p < 0.05). Furthermore, hospitals that faced more competition from geographically close competitors showed a more pronounced improvement of CQI-scores 0.004 to 0.05 than other hospitals (p < 0.001). Conclusion Our results show that patients reported improved experiences measured by the CQI between 2006 and 2009. CQI levels improve at a faster rate in areas with higher levels of competition. Hospitals confronted with forced public publication of their CQI responded by enhancing the experiences of their patients.


Background
In the last two decades, several Western countries introduced some form of managed competition in their health care system [1,2]. Common goal of these reforms is creating a demand driven system that provides more patient centered care [3]. To achieve this goal the quality of health care providers needs to be assessed and publicly reported [4,5]. Patients and health plans may then use quality information to make informed choices between health care providers.
The public reporting of provider quality can stimulate quality improvement through informed patient choice, quality contracting of providers by health plans and/or by intrinsic motivation of health care providers [6]. Previous studies have shown that public reporting of quality information does stimulate hospitals to initiate quality improvement projects. It remains unclear if-and to what extentthe introduction of managed competition combined with public quality reporting may affect quality [7][8][9].
We set out to answer this question in the context of the Dutch 2006 Health Insurance Act (HIA) reform. It was enacted to stimulate managed competition in the hospital market. Before the reforms all hospitals were financed through 'input' reimbursement. Hence hospital budget did not directly depend on production, but rather on historic agreements with the insurer regarding their budget. The after reform characteristics of the Dutch health can be summarized as followed: [2,10] • There is a mandatory basic health insurance for everyone, purchased through private insurers (both profit and not-for-profit are possible); • There is annual consumer choice of insurer and insurer products and choice between in kind and restitution policy; • Insurers are obliged to have a system of open enrollment and are compensated via a risk-equalization system; • There is a mandatory deductible of € 150 and avoluntary deductible up to € 650 per person per year; • Insurers are expected to contract selectively with competing health care providers; • The government requires health care providers to release quality information.
To enable provider competition, products were defined in terms of Diagnosis Treatment Combinations (DTCs), which are to some extent comparable to diagnosis related groups [11]. A number of initiatives were aimed at making provider quality transparent. These include widespread measurement of medical quality indicators that aim to indicate outcome utility and the measurement of the Consumer Quality Index (CQI) that aims to measure process utility. It is partly based on the Consumer Assessment of Healthcare Providers and Systems [12]. The CQI therefore is a partial measure of healthcare quality based on more objective consumer experiences rather than a more subjective satisfaction [13].
From 2006 onwards health plans were increasingly stimulated to negotiate with hospitals on price, quantity and quality of care. While in 2006 negotiations were restricted to products accounting for 7% of the total hospital budget, this was increased to 34% in 2009, and is expected to further increase to 70% in the near future. Consequently, hospitals are increasingly exposed to competition [14]. However, the bargaining power of health plans remains limited due to a limited use of selective hospital contracting by health plans. In addition health plans are currently reluctant with proactive member channeling toward preferred hospitals [15]. While this may have reduced competition during 2006-2009, the threat of significant competition in the future may have sparked hospital policy changes.
We examined whether the patient experiences of hospital care improved in the period 2006-2009. In addition we investigated whether forced public reporting of hospitals and higher levels of competition-in line with previous studies and policy objectives-were associated with better patient experiences [7,9,16,17].

Available data
Hospital care experiences of patients were measured in 2006, 2007 and 2009. Each year the questionnaire was sent to a random sample of patients who stayed at the hospital during the respective year. In 2006 31 hospitals out of total 94 Dutch hospitals participated in the measurement. This number increased to 78 in 2007 and 93 in 2009 [13].
Under Dutch law no specific ethical approval was required for our research as the respondents were informed during filling in the questionnaire that their answers may be used for research purposes.
The CQI consisted of seven quality aspects (Physician communication, Nurses communication, Pain treatment, New medication communication, Accommodation, Discharge information and Nurse services) in 2006. In 2007 eight quality aspects were added and one was deleted (see Table 1). In 2009 one quality aspect was deleted (see Table 1). All quality aspects were rated based on multiple answers which were scored at a four point scale. This scale varies between one for the lowest possible score and four indicating the highest possible score [13].
Not all quality aspects were suitable for our analysis. The scale of the quality aspect 'information after discharge' was adjusted in 2007 and therefore lacks comparability between the years 2006 and later years. In the quality aspects New medication communication, Accommodation and Intake conversation, questions were deleted after analyses showed they did not discriminate between respondents. Consequently trends over time may be biased and these quality aspects were (partly) excluded from further analyses (see Table 1) [13].

Primary analyses and publication
Each year the results for the individual hospitals were case mix corrected using multilevel linear regression analyses (consumer experiences were nested within hospitals) [18]. Means with comparison intervals (1.39 standard error) were calculated per quality aspect and per hospital adjusted for age, education and reported physical health, sex and reported mental wellbeing [19]. Next, hospitals were divided in three groups and given a 'star rating': one star if their interval was completely below the average across all providers, two stars if their interval overlaps with the average across all providers and three stars if their interval was completely above the average across all providers, so that the public could easily interpret the results of the individual hospitals.
Depending on the dominant health plan in the catchment area of the hospital, 16 out of 31 hospitals were confronted with forced publication their 2006 results. This decision was made by the health plan without consultation of hospitals. In 2007 none of the results were publicly reported. All 2009 results will be published at the health care portal http://www.kiesbeter.nl and some results are already publicly reported at health plans' websites. All hospitals have received their CQI results for each of the years [20].

Secondary statistical analyses
We combined the data of the three years and conducted a multilevel linear regression analyses per quality aspect We added dummies for groups of hospitals with different participation patterns to correct for different CQI scores between these groups of hospitals. To assess if hospitals which were confronted with published CQI results by health plans in 2006 outperformed hospitals which were not, we included a dummy variable for those hospitals which we included in our multi level linear regression model. In order to identify hospitals that faced more competition from neighboring hospital we used the Herfindahl Hirschman Index (HHI) of hospitals. This HHI was used as an explanatory variable in an ordinary least square (OLS) regression, while correcting for time effects and case mix variables [21]. All analyses were conducted in Stata/IC version 11.0.  Table 2 shows the non-response analyses, in 2009 respondents were significantly younger than non-respondents.   Publication of results consequently has a non-significant consistent positive relationship with higher performance for 10 quality aspects. The quality aspects show significant improvement 'Physician communication' and 'Accommodation' at the p < 0.05 and P < 0.01 level. Table 5 shows that the combined analyses of quality aspects yields consistent results: in 2009 hospitals improved their performance with 0.034 (p < 0.05) to 0.060 (p < 0.01) points compared to 2006 and 2009. Publication of results is associated with an improved the performance of hospitals of 0.027 (p < 0.01) to 0.030 (p < 0.05) points.

Performance and level of competition
The aggregated analyses of quality aspects 1-3 and 4-12 shows that the HHI is inversely significantly related to the performance of hospitals at the p < 0.001 level, meaning that hospitals who face a higher level of competition outperform hospitals who face a lower level of competition, as shown in Table 6.
The coefficient is small, but this is largely caused by the scale of the HHI, which theoretically ranges from 1 to 10,000, but in our dataset ranges from 807 to 9,133. If one multiplies coefficients by these differences, the potential impact of the HHI is equal to a coefficient of approximately 0.004 to 0.05 points on the CQI which varies in our dataset between 2.97 and 3.88. Thereby this effect is in the same order of magnitude as the found time effects in Table 5.

Discussion
We have examined the effects of the introduced transparency and competition between providers on patient experiences within the Dutch hospital market. This study is the first which systematically evaluates these effects within the Dutch setting. In addition we were able to use a dataset which enabled us to compare hospitals which were confronted with transparency early on with hospitals which were not. Our findings are as follows: First, we observe that patient experiences measured with the CQI improve over the course of years. Secondly, hospitals which are confronted with forced transparency early on are improving faster than hospitals which were not. This is evidence that the   [22]. Thirdly, we find that higher levels of competition adjusted for case mix differences and time effects are related with better patient experiences,.
Although the findings in our study are in line with previous studies our study may have suffered from possible limitations [8,23]. First, in this study we show that Dutch hospitals are improving their performance, but they found improvements are fairly small. It is well documented that CQI differences between good and bad performing providers are in the same orders of magnitude as we have found [24,25]. Therefore we believe that the differences in patient experiences between providers are meaningful and relevant for policy makers, health plans and hospitals. Moreover, we believe that current levels of competition are weak and our effects are likely to become stronger as competition increases. Secondly, and in line with our hypotheses and previous studies, our study indicates that a higher level of competition is related to better performance of hospitals. The effect of a higher level of competition on patient experiences is in the same order of magnitude as the found time effects [8,16]. We use the HHI to indicate the level of competition that is calculated in 2004. Although some mergers took place within the Dutch hospital market since 2004, changes in market share over time are fairly modest [26]. In addition, most mergers took place on a board level, while merging the enterprises of the medical specialists on the multiple locations often lags several years behind. This maintains the same pre-merger level of competition between medical specialists. Therefore we believe that adjusting the HHI over the course of years would not have altered our conclusions.

Conclusions
In conclusion, our results show that patients reported improved experiences measured by the CQI between 2006 and 2009. CQI levels improve at a faster rate in areas with higher levels of competition. Hospitals confronted with forced public publication of their CQI responded by enhancing the experiences of their patients.

Additional material
Additional file 1: Supplementary data. Coefficient of combined quality aspects HHI Quality aspects 1-3# **** -5.42 10 -6 *** Quality aspects 4-12 -4.89 10 -6 *** *Significant at p < 0.05, **significant at p < 0.01, ***significant at p < 0.001 **** Given the skewed distribution of the error term we estimated confidence intervals we applied a non parametric bootstrap procedure with 10,000 replications and hospital stratified sampling. Confidence intervals were based on bias corrected percentile score and the reported coefficients are within these confidence intervals, meaning that the found coefficients are significant. For additional analysis regarding potential outliers that may have influenced our regression results, please see Additional file 1. This analysis shows that this is not the case.