Design And Effects Of Rural Chronic Care Delivery Integration System From Perspective Of Complex System Theory

[Purpose] The fragmented rural health system had impeded the progress of forging the pieceful service delivery into an integrated one ever since the New Rural Cooperative Medical Schema had initiated. With the need of rural chronic patients becomes more complex and integrated, the system had no longer echoed to the rural chronic patients. Not only would this result in the devastation of the quality of care, but also raised up the disease burden and hampered the system efficiency. While integrated care has been proved a useful way of improving the patient health by providing care continuously and coordinately, the exact service model and its motivation model were not clear in rural areas where the human resources were urgently in need. To solve this problem, we searched relative literatures and investigated some rural areas, and had finally decided the intervention package to be a combination of a Multi-Disciplinary Team (MDT), the Multi-Institutional Pathway (MIP) and the System Global Budget and Performance-based Payment (SGB-P4P) reform. With this plan, we hope to find an answer to the above question.To gather solid evidence, we initiated a multi-faceted intervention in Qianjiang, Chongqing municipality--southwestern part of Chinese rural area from July2012to December2014, with the aim to discover the effectiveness of the potential intervention mode. After half year’s pilot and2years’ following up, we had collected adequate evidence from both patients and suppliers. Under corresponding statistical analytical plan and process evaluation, we had fully finished the impact evaluation of this intervention program. And with comparison to other similar practice home and abroad, we had found out not only the effects but also the effecting mechanisms and the pre-requisites of health system integration. Giving out the following results and conclusions, we are hoping to provide an evidence-based decision making process to the targeted policy makers and other researchers too.[Methods] The study adopted a prospective cohort study in rural communities about chronic care.6towns were selected and grouped randomly into3groups from24rural towns, and1245hypertensive patients and327diabetes patients were sampled with certain inclusion and exclusion criteria.1425hypertensive and286diabetic patients were followed at last. Treatment group1(TR1) was intervened with MDT and MIP; treatment group2(TR2) was intervened with both SGB-P4P and MDT and MIP; the control group (CT) was a blank controller. We mainly used difference-in-difference (DID) and difference-in-difference-in-difference (DDD) model to examine the trial effectiveness between groups, and used repeated measurement ANOVA, linear regression, logistic regression, chi-square, ANOVA and other suitable statistical methods to analyze the data. The primary outcomes were variables such as blood pressure, quality of life, drug compliance, life styles, continuity of care, integration degree of MDT and the suppliers. We had assumed that, under the help of Complex System theory and Theory of Change, we could build the relationship between perspective payment reform/integrated service model and the health behaviors/patient health outcomes, through gradually improving the continuity of care and patient awareness, knowledge and behavior. The primary assumption was that:the treatment groups would perform better on each of the indexes compared to the control group. The second assumption was that:TR2would perform better on each of the indexes compared to TR1.[Results] Control variables:The background variables of the township hospitals were found to be insignificantly different between3groups both before and after the intervention, such as the population, the average income, and the economic income, reach and capacity of the township hospital. Also, the background variables of the patients were found mostly to be insignificantly different between3groups both before and after the intervention, such as gender structure, age, family structure and availability to village clinics. Except for the following indexes had shown significant differences in the baseline study:the education level, the availability to the township hospital and the county hospital. As a result, the above variables would be included in the DID regression model at a complementary part for explanation. The revised DID model were given in the main text.Outcome variables:Take hypertensive patients as an example, we found the main outcomes to be as follows:The first part were outcomes from the patient health.1.1Blood pressure and control rate of BP. The BP were measured16times on1245patients during the process of follow up.4were taken before the intervention and12were taken after.Before the intervention, the changing tendency of2TRs were tested to be the same, but with a difference in the BP level:TR1were measured to be150.6mmHg and TR2to be148.2mmHg (P=0.002). The control group was tested to have a significantly different cure with2TRs (P<0.001), and the average BP level was also different:139.8mmHg (P<0.001). Therefore, we had concluded that it was basically comparable between the2TRs, and we also needed to control the difference from CT in the regression model.After the intervention, the BP level has decreased by6.7%(P<0.001) and5.9%(P<0.001) separately in TR1and TR2. The CT had shown no change (P=0.229). And the repeated measurement ANOVA test had shown the different change tendency with time in these3groups (P<0.001). By DID regression, we had found out the outcomes to be more exaggerated, the BP level were adjusted to have decreased by7.6%and6.7%in TR1and TR2.After intervention, the control rate of BP has increased by36.9%(P<0.001) and27.6%(P<0.001) in TR1and TR2compared to CT. between2TRs, there also showed a significant difference (P=0.001).1.2Quality of life. The QOL were measured twice, once at the baseline among1425hypertensive patients and the other at the end-line among1245. We used simplified SF-36scales, and found out that, compared to the CT, score of TR1had raised by15.5%(P<0.001), and the score of TR2had raised by8.2%(P=0.032). Between the2TRs, there was no difference (P=0.083).The second part were outcomes from patient behaviors. The data were collected in the same way as the QOL.2.1Drug compliance. We used simplified Morisky scale to measure the DC, and we mainly used chi-square to test the group differences both between and within3groups. We had found out as follows: Firstly, there were significant differences both before and after intervention (P=0.004); Secondly, the proportion of patients with good drug compliance had increased from41.2%to45.1%(P=0.380) in TR1, and the same proportion had decreased by40.5%to39.2%(P=0.741) in TR2. The decreasing tendency showed no significant difference in CT, from52.5%to51.3%(P=0.791).2.2Life styles. We mainly used chi-square and Bonferroni adjustment to make the pairwise comparison of salt-controlling and fat-controlling behaviors between groups both before and after intervention. The results were as follows:Firstly, there were no differences among3groups on salt-controlling (P=0.538) and fat-controlling (P=0.775) behaviors before intervention started;Secondly,after intervention,the salt-controlling behaviors had showed better performance both in TR1(P<0.001) and TR2(P<0.001); but still there was no difference between2TRs (P=0.693);Thirdly, after intervention, the fat-controlling behaviors had showed better performance both in TR1(P<0.001) and TR2(P=0.024); but still there was no difference between2TRs (P=0.165).2.3Choice of first diagnosis. Examining the choice of township hospitals or county hospitals as the first choice of diagnosis, we had found out that:Before the intervention, more patients chose to seek higher institutions to have their treatmenet in CT (P<0.001), and the situation remained the same afterwards (P<0.001);The choice of first diagnosis showed no differences in TRs before intervention (P=0.508), however after intervention, the patients shifted to lower institutions as their first choice in TR1(P<0.001).Other secondary outcomes were given in the main text, including:the self-health-awareness, self-knowledge-awareness,self-efficacy, disease burden, satisfaction and trusts towards township doctors.The third part were outcomes from supplier behaviors.We used questionnaires of D’Amour collaboration degree of team and professionals to measure the integration degree of MDT and its members; we used team reports and the medical records to evaluate the team cooperation behavior and continuity of clinical services between two level-hospitalization. The measurements were taken once each both at the baseline and end-line study.3.1Continuity of two-level hospitalization service. The data were collected from medical records both in the township hospitals and county hospitals where the in-hospitalization had taken places consequently within30days for the same patients treating the same or relative diseases. The study selected records from December2011to June2014,279suits were included. The continuity of services were evaluated from5aspects, the enquiry of county doctors, the appropriateness of township hospital treatment, the continuity of two-level treatment, the downward referral and the entire continuous behavior.3.1.1To explore the factors of continuity of two-level service, we used DDD model to test the influences of the following factors:group, intervention time and referral manner. Observing the significance of coefficients, we had found out that:Firstly, compared to TR1, the COC in TR2had increased by26.9%(P=0.032);Secondly, compared to the non-continuous referral manner, the COC had increased by36.7%(P<0.001);Thirdly, compared to before, the COC after intervention had increased by10.4%(P<0.001)；Compared to the enquiry behavior (P=0.415), the appropreateness had more significantinfluence (P<0.001).However, there were no interactions within group, intervention time and referral manner.3.1.2To explore the group differences of5aspects of COC after intervention, we used the multi-regression.Firstly, the TR2had performed better on the enquiry (P=0.037), downward referral (P=0.034) and entire COC (P=0.005); and the TR1had performed better than CT, and the significant level on each of the behaviors was P=0.042,P<0.001and P=0.039;Secondly, the TR2had performed better on the appropriateness (P=0.039) and two-level COC (P=0.034); however the TR1had shown no different performance on these2aspects compare to CT, and the significant level on each of the behaviors was P=0.813, P=0.928;Thirdly,on the enquiry, appropriateness, two-level COC and entire COC, the continuously referred patients had shown better performance than the non-continuously referred patients, and the significant level on each of the behaviors was P=0.009, P=0.003, P=0.002, P=0.003.3.1.3To explore the differences of the entire COC both before and after intervention, we used DDD model to test the interactions terms of the3potential influential factors.Firstly, compared to the CT, the TRs had shown increase in entire COC by19.2%(P<0.001) after intervention; and the continuously-referred patients showed11.0%higher (P=0.037) than non-continuously referred patients in TRs;Secondly, compared to TR1, the entire COC had shown an increase by33.8%(P<0.001) after intervention in TR2; and the continuously-referred patients in TR2showed15.0%higher (P=0.119) than non-continuously referred patients.3.2Team cooperation behaviors. The data were collected from team reports for16times at the interval of every2months. Before intervention, the performances were basically similar among3groups within the recall period (6months). We mainly used repeated measurement ANOVA and S-N-K Q test to make the pairwise comparisons.The outcomes variables were:upward referrals within town (P<0.001), downward referrals within town (P<0.001), upward referrals within county (P=0.003), downward referrals within county(P=0.016), communications between county-township doctors (P<0.001), communications between clinical and primary care givers within township hospital (P=0.015), communications between suppliers and patients(P=0.965), and accompanied referrals (P=0.003).Except for "communications between suppliers and patients", TR2performed significantly better than TR1, and then CT.3.3Team integration degree.264questionnaires were collected by adapted D’Amour Team Questionnaire. The measuring aspects were:the team operation, the common goal, the mutual acknowledgment, the integrated service model and the shared interests.Firstly, by using rank sum test and S-N-K Q test, we found out that within each aspect, the ranking was the same:CT<TR1<TR2(P<0.001);Secondly, by using multi-regression, we found out that the integration degree varied with identities of doctors:the county hospital professional had shown a lower scoring than both township doctors and primary care givers (P=0.019). Therefore, in terms of team integration degree, the CT could be recognized as the Potential Cooperation, the TR1the Developing Cooperation and TR2the Positive Cooperation.3.4Professional collaboration degree.120doctors were investigated both before and after intervention. The outcome variables were measured from6aspects:the willing, the motives, the mutual trust, the familiarity, the capacity and the intensity of collaboration. And we mainly used the ranking sum test and multi-regression, the results were as belows:Firstly, each aspect of scores ranked mostly at low level and middle level, seldom high level;Secondly, the entire collaboration degree showed higher rank after intervention and also in the TR2;Thirdly, the scores of collaboration aspects showed different rankings among different professionals. For detail, please see the main text.4. Process evaluation.Firstly, the implementation degree. By observing and analyzing the processes of exposure, enrollment, reach and following up, we had found out the implementation degree to be only61%during the actual intervention; among which, the patient were reached100%, the supplier72%, and the hospitals85%.Secondly, the contamination. We had also found out that there were possible chances of John Henry Effects in the TR1and thus group positive contamination was likely to happen, which potentially decreased the significant level of TR2.[Conclusion]1. The intervention mode of MDT&MIP had been proved to have a positive influence on BP control rate; and the effect would be exaggerated with company reform of SGB-P4P;2. The intervention mode of MDT&MIP had been proved to have positive influences on COC, team cooperation behavior, team integration degree and professional collaboration; and the effects would be exaggerated with company reform of SGB-P4P;3. The intervention mode of MDT&MIP had been proved to might have positive influences on QOL, drug compliance and life styles after adjusted regression, and have a less effects on self-awareness, self-efficacy, disease burden relief and satisfaction; however under these terms, there were no proof of marginal help from SGB-P4P.To make a conclusion, the integrated care had been proved to be effective in improving the continuity of care, and then the patient health status; the perspective payment reform had been proved to be helpful in reinforcing the collaboration and cooperation among different suppliers, but with no evidence to add help in terms of patient health status.[Advantages&Limitations] Advantages:Firstly, the health policy decisions were seldom made based on rigid evidence-based design, and this study was the first of its kind to testify the effectiveness of integrated care in Chinese rural areas. Secondly, the study developed an intervention package as the solution plan to overcome the shortcomings of the inadequate human resources and behavior instructions, also,at the same time, the study had introduced the System Global Budgets and Performance-based Payment reform to fully motivate all suppliers to work as a whole and towards patient good. Thirdly, by all means, this study provided a trust design and proof of the above intervention. Limitations:The study has its limitations. Firstly, there was gap between the ideal intervention and actual implementation; secondly, the indexes lack adequate specificity and with low discrimination. Thirdly, although we used Theory of Change as a basic theory to guide the study proving, however there were still room for improvement in the statistical analysis methods and further exploration in this field are needed.