Five-Year Impact of a Continuous Quality Improvement Effort Implemented by a Network of Diabetes Outpatient Clinics Free

Starting from 2000, a total of 22 diabetes outpatient clinics in Sicily agreed to adopt the same electronic medical record system (EuroTouch) for the everyday management of outpatients to facilitate the comparative evaluation of their performance. In 2001, a meeting with all the participants was held to identify a set of process and intermediate outcomes indicators that was comprehensive yet parsimonious in order to minimize the burden of data collection. Existing quality indicators have been reviewed, with particular attention to the work done by the National Diabetes Quality Improvement Alliance in the U.S. (13). All the measures identified were evaluated for their importance, scientific soundness, and feasibility. Required criteria included 1) credible evidence linking process measures to important clinical outcomes and modifiability of the clinical outcome measures (i.e., they could be improved) by the efforts and interventions of health care systems; 2) feasibility (i.e., whether the measure could be collected accurately, reliably, and at a reasonable cost); and 3) variability across health care settings, so that there would be opportunity for improvement. As for diabetes complications, while their importance and scientific soundness as process (i.e., eye examination, foot examination) and outcome measures goes uncontested, their definition was not sufficiently standardized to allow between-center comparisons; in fact, open text was generally used to describe the monitoring, presence, and severity of the complications, hampering the ability to extract the necessary information from electronic medical records.

Specific software was then developed, enabling the extraction of the information needed for the profiling of quality of care. Data were centrally analyzed anonymously, while each individual center had the possibility of obtaining the information on its performance directly from the electronic record system, using specific queries.

Data were analyzed annually, from 2001 to 2005, and results were discussed once a year in ad hoc meetings involving the head of each of the participating diabetes clinics. On those occasions, the performance of the different centers was ranked against the “best performers,” the reasons for variation were openly discussed, and the possible solutions were fully evaluated. In all the steps of the process, anonymity was ensured. Following the plenary meetings, data were provided to each clinic to allow internal discussion, led by the head of the clinic, and involving all the personnel practicing in the structure (physicians, nurses, and dietitians).

The analyses of the data included all the patients aged ≥18 years who had had at least one encounter with the clinic during the index calendar year. Process measures included frequency of measurement of A1C, blood pressure, lipid profile (total and HDL cholesterol and triglycerides), and microalbuminuria. Process measures were expressed as percentages of patients monitored at least once during the previous 12 months. Intermediate outcome measures included mean A1C, blood pressure, lipid profile, and BMI. In case of multiple records during the year, the last value was considered for the analyses. LDL cholesterol was estimated by the Friedwald equation. For selected outcomes, we also considered the proportion of patients with satisfactory values as well as the percentage of those with unacceptably high values. Outcomes were considered satisfactory if A1C levels were ≤7.0%, blood pressure values were ≤130/85 mmHg, and LDL cholesterol levels were ≤100 mg/dl. Unsatisfactory outcomes included A1C levels ≥8%, blood pressure values ≥140/90 mmHg, and LDL levels ≥130 mg/dl.

Finally, the percentage of use of specific classes of drugs was calculated, including diabetes drugs (oral agents, insulin, and insulin plus oral agents), antihypertensive drugs (ACE inhibitors, angiotensin receptor blockers [ARBs], diuretics, and calcium channel blockers), statins, and aspirin. The last prescription of the aforementioned drugs during the year was considered. The percentage of use of antihypertensive drugs was calculated taking the number of patients with hypertension as the denominator. The percentage of use of ACE inhibitors or ARBs was also calculated in the subgroup of patients with microalbuminuria/macroalbuminuria, defined as an albumin excretion rate >20 mg/l on morning urine spot. Aspirin use was evaluated in individuals aged ≥40 years. Data are summarized as means ± SD for continuous variables and percentages with 95% CIs for proportions. Information on variation for process and outcomes measures was displayed graphically using box-plots. Using these graphs, each center was able to locate its own performance with respect to the overall picture.

From January 2001 to November 2005, a total of 26,782 patients have been seen at least once in the participating clinics; on average 12,000 patients have attended the centers every year (range 9.647–14.247). The mean age of the study population was 65 ± 12 years, while the proportion of males was of 49%. Rates of individuals who received processes of care for diabetes and drug prescriptions are reported in Table 1, while intermediate outcomes attained are reported in Table 2.

The proportion of patients with at least one measurement of A1C constantly increased across the years, from 56.6% in 2001 to 76.1% in 2005 (Table 1). At the same time, the between-center variability in A1C monitoring decreased, as shown by the decreasing length of the boxes in Fig. 1A. Mean A1C levels also progressively decreased from 7.8 ± 1.7 in 2001 to 7.3 ± 1.5 in 2005 (Table 2). Between- center variability in A1C levels tended to decrease along the years as well (Fig. 1B). When looking at the proportion of patients reaching specific therapeutic targets, it emerged that the percentage of individuals with A1C values ≤7.0% increased by >15% (from 43.3 to 59.9%) during 5 years of observation (Table 2). These positive changes in metabolic control were associated with a progressive decrease in the percentage of individuals on diet alone (from 11.9% in 2001 to 6.5% in 2005). In parallel, the proportion of patients treated with oral agents increased from 52.8 to 58.2%, while no major changes were detected in the use of insulin, alone or in association (Table 1).

The proportion of individuals with at least one blood pressure reading during the year only slightly increased over time (Table 1) and was associated with a decrease in between-center variability (Fig. 1C). Small declines in mean systolic and diastolic blood pressure values were also documented and were associated with an increasing rate of individuals with values on target (i.e., ≤130/85 mmHg), growing from 37.8% in 2001 to 48.5% in 2005 (Table 2). Between-center variation in blood pressure levels was very small and did not substantially change during the study period (Fig. 1D).

Overall, the prevalence of individuals with hypertension slightly increased over the years (from 56.7% in 2001 to 68.4% in 2005), probably as a reflection of better reporting. The percentage of patients with hypertension treated with antihypertensive drugs substantially increased from 22.4% in 2001 to 72.9% in 2005 (Table 1). Increasing rates of use of all antihypertensive agents were documented; the proportion of individuals treated with two or more drugs also raised considerably across the years (Table 1).

Monitoring of lipid profile showed a constant percentage increase over time (Table 1) and was associated with a reduction in between-center variability in rates of lipid testing (Fig. 1E). As for intermediate outcomes, a marked decrease in total cholesterol and LDL cholesterol levels was documented, while triglycerides and HDL cholesterol levels showed only marginal temporal variations. The proportion of individuals with LDL cholesterol levels on target also remarkably increased from 19.4 to 44.1% (Table 2).

The improvement in LDL levels did not translate into a reduction in between- center variability (Fig. 1F). This finding can be explained by the fact that within a general trend of improvement in mean LDL cholesterol levels, some clinics attained a markedly better result than others, thus increasing the between-center variability. The positive changes in total and LDL cholesterol levels were associated with a sharp increase in the use of statins, growing from 4.1% in 2001 to 27.5% in 2005. Among patients receiving statins, the percentage of those on target in 2005 was 49.0% (46.9–51.1), while 26.5% (24.7–28.3) still had values ≥130 mg/dl. Similarly, among individuals not receiving statins only 41.7% (40.3–41.1) showed LDL levels <100 mg/dl, while 26.5% (24.7–28.3) were not treated despite LDL values ≥130 mg/dl.

Monitoring of microalbuminuria, while showing an important trend of increase over the years, was still performed in only one-third of the patients in 2005 (Table 1). Among individuals who had received the test, the proportion of those with macroalbuminuria progressively decreased from 30.5% in 2001 to 20.5% in 2005 (Table 2). The use of ACE inhibitors or ARBs in individuals with macroalbuminuria increased from 35.9 to 44.1% (Table 1). Finally, mean BMI values remained unchanged, while the use of antiplatelet agents in individuals over 40 years raised from 5.8 to 27.4% (Table 2).

Our study shows that sharing an electronic health record system has a significant potential for conducting practice-based quality-of-care studies across large numbers of outpatient practices. This was a preliminary, fundamental step for reaching a consensus in how to measure the quality of care in priority areas, promote critical evaluation of current practice, develop process improvements, and reduce practice variation.

By adopting standardized process and outcomes measures, the study was able to document a tangible improvement in the quality of diabetes care provided by outpatient clinics in Sicily over 5 years. Recent quality improvement initiatives have consistently documented an improvement in processes of diabetes care, without a corresponding increase in treatment intensity (10) or an improvement in intermediate outcomes such as A1C levels (9–11), blood pressure control (9,10), or LDL levels (9,10). In our study, both process and intermediate outcome measures constantly improved over time, and the link between process and outcomes improvements is strongly supported by the information relative to intensity of treatment. One can in principle hypothesize that some of the improvements documented, particularly for process measures, can be attributable to a more complete reporting of the information in the database during the years. Nevertheless, we found a strong consistency between increasing rates of monitoring, increasing drug prescription, and better levels of intermediate outcomes. This was particularly evident for the link between increased lipid monitoring, increased prescription of statins, and improved total and LDL cholesterol levels. That this is not an unspecific finding is further confirmed by the observation that triglycerides and HDL cholesterol levels, less affected by statin treatment, only marginally changed across the years. A similar, though less marked, improvement was documented for blood pressure monitoring, treatment, and control. Improvement in blood pressure levels and targets achievement was less prominent notwithstanding a remarkable increase in use of antihypertensive agents. Improved reporting of medication prescription or suboptimal treatment are likely explanations for our findings. The need for more aggressive therapeutic approaches is further suggested by the fact that <50% of individuals with hypertension were treated with two or more antihypertensive agents.

As for metabolic control, improvements in monitoring and A1C levels were particularly evident, but they were associated with only a moderate increase in the proportion of individuals treated with oral agents. The use of higher doses of oral agents or insulin might concur in explaining our findings, though we were unable to collect information on drug doses.

Overall, the temporal changes documented seem more pronounced than those documented in the U.S. over 10 years (8), particularly for metabolic control, both in terms of mean A1C levels and proportion of patients with A1C levels <7.0%. Furthermore, while no changes in blood pressure levels were documented across the years in the U.S., the proportion of individuals at goal (i.e., ≤130/85 mmHg) constantly increased in our study.

Another important piece of information derived from our study is represented by the reduction in practice variation, which is generally considered a major contributing source of inappropriate care as well as of resource consumption (14). We initially found a remarkable variation across centers, suggesting gross differences in the ability of specialist structures to provide adequate care for people with diabetes. After 5 years from the launch of the initiative, such a variability was substantially reduced for many processes and some of the intermediate outcome measures.

Overall, our results support the concepts that direct measurement, feedback, and reporting of intermediate outcome levels or of level of medication management may enhance the effectiveness of care (9). A key feature of the continuous quality improvement effort implemented in Sicily is represented by the decision to use the “best performers” approach (15). In other words, clinicians were not faced with theoretical standards, often perceived as unrealistic in their structural and organizational setting, but rather with the performance of centers operating in the same geographic area under similar conditions. By comparing their own performance with that of centers reaching better overall results, specialists could easily realize the real margin of improvement made possible by simply increasing the level of attention to disease monitoring and treatment.

Despite the satisfactory achievements, a substantial room for improvement in the care of diabetes still persists. One in three patients still has A1C levels ≥8% or LDL cholesterol levels ≥130 mg/dl, while one in two has blood pressure levels ≥140/90 mmHg. This situation is mirrored by the persistence of an elevated percentage of individuals not treated with statins or multiple antihypertensive agents, despite elevated LDL cholesterol and blood pressure levels. On the same line, monitoring of microalbuminuria is still unsatisfactory, and among individuals with microalbuminuria, ACE inhibitors and ARBs are underutilized. Getting more patients to goal thus represents an important priority of the initiative in the years to come. The inclusion of additional indicators representing broader aspects of diabetes care (i.e., eye and foot examination, education, influenza vaccine) as well as the addition of distal outcomes (i.e., cardiovascular events, severity of retinopathy) also constitute a necessary step to implement. Nevertheless, the results obtained so far, if maintained, would translate into a substantial reduction in the risk of major complications.

Our study has some limitations. First, this is not a comparative study. Therefore, we cannot exclude the possibility that the positive changes in quality of care documented over 5 years are at least in part the results of a historical trend rather than a specific effect of the quality improvement initiative. Nevertheless, substantial achievements have been obtained in a relatively short period of time, thus suggesting a more specific effect of the program. Second, the considerable success documented was obtained without allocation of extra resources or financial incentives but simply through a physician-led effort made possible by the commitment of the specialists involved. While this is a qualifying aspect of the initiative, it can at the same time represent a factor that might limit its generalizability to other areas where clinicians do not display a similar willingness to share their experiences with colleagues.

In conclusion, the experience of the diabetes centers in Sicily opens important perspectives. Over 300 diabetes outpatient clinics throughout Italy now share the same electronic health record system, and the list of process, intermediate outcomes, and treatment intensity indicators have been endorsed by the Associazione Medici Diabetologi. Starting from 2006, a total of 90 centers have already agreed to implement the same process realized in Sicily, and information on over 120,000 patients has been collected for the calendar years 2005 and 2006. It will be a major challenge to show that the same results obtained in a restricted, rather homogeneous, area can be replicated on a much larger scale. If so, substantial benefits can be foreseen for individuals with diabetes in Italy in the years to come.

Gioacchino Allotta, Trapani; Gianclaudio Allegra, Palermo; Gaspare Cordaro, Ramacca (CT); Francesco D’Agati, Palermo; Antonino Di Benedetto, Messina; Maurizio Di Mauro, Catania; Marina Fulantelli, Palermo; Arcangela Garofalo, Vittoria (RG); Calogero Giacchetto, Caltanissetta; Enrico Lisi, Tremestieri (CT); Nino Lo Presti, Marsala (TP); Ignazio Lorenti, Lentini (SR); Mario Manunta, Palermo; Giuseppe Mattina, Palermo; Barbara Mendola, Caltagirone (CT); Pietro Pata, Messina; Fabio Pellegrini, S. Maria Imbaro (CH); Giuseppe Reina, Adrano (CT); Giovanni Ridola, Palermo; Filippo Runello, Acireale (CT); Giovanni Saitta, Messina; Giuseppe Sanfilippo, Giarre (CT); Maria Vaccaro, Palermo; Antonio Nicolucci, S. Maria Imbaro (CH).

The study was supported by LifeScan Italia.