Landscapes on Prevention Quality Indicators: A Spatial Analysis of Diabetes Preventable Hospitalizations in Portugal (2016–2017)

Preventable hospitalizations due to complications of diabetes mellitus (DM), represented by the related prevention quality indicators (PQI), are ambulatory care-sensitive conditions that can be prevented and controlled through effective primary health care (PHC) treatment. It is important to reduce mortality and promote the quality of life to diabetic patients in regions with higher hospitalization rates. The study aims to analyze the results of the DM age-sex-adjusted PQI, by groups of health centers (ACES), distributed in the Portuguese territory. The most representative PQI at a national level were identified, and the trends were mapped and analyzed. Also, it presents the ACES with the highest age-adjusted rates of avoidable hospitalizations for DM. The absolute number of preventable hospitalizations for all DM complications in Portugal has decreased by 20%, thus passing from the rate of 79 in 2016 to 65.2/100,000 inhabitants in 2017. Despite the improvement in results for PQI 03, 20 of 48 ACES that were above the national 2017 median rate in 2016, achieved better results the following year, and for the overall preventable diabetes hospitalizations (PQI 93) only 11 out 39, revealing the need for further studies and PHC actions to improve the diabetic quality of life.


Introduction
Diabetes Mellitus (DM) is certainly one of the non-communicable diseases that represent a major challenge in global health in recent decades, counting for some 1.6 million deaths annually [1,2]. In Portugal, according to the World Health Organization data (WHO), DM affected 13.3% of the population between 20-70 years old and was responsible for 5% of all deaths. The national health ministry promotes the monitoring and prevention of DM through of the National Diabetes Prevention and Control Program [3,4]. The promotion of integrated care for diabetic patients in the system has been evolving since 2006, with horizontal integration of the health and social assistance sectors, through the creation of the long-term care network of service providers. Integrated Diabetes Units were created to manage better the complexities of this condition, a service provided by specialists. The monitoring

Study Design and Variables
This study is a retrospective observational analysis of secondary data from the Portuguese national hospital databases for diabetes avoidable hospitalizations.

Measurement
The prevention quality indicators have been computed according to AHRQ specifications, using the inpatient hospital data available through the national hospital morbidity database, which contains inpatient and outpatient hospitalizations data occurring in all public hospitals in mainland Portugal to determine the numerator. The Portuguese public hospitals account for 79% of all inpatient discharges and represent 72% of all inpatient hospital beds [21] The information regarding diagnoses and medical procedures is coded according to the International Classification of Diseases (ICD-9-CM and ICD-10-CM/PCS). The only inpatient discharges data covering the years 2016 and 2017 were used for this study, since these datasets are the latest available. The 2011 population census was used to determine the PQI denominators. Direct standardization was implemented to obtain age-sex-adjusted PQI using the 2011 Portugal census population as the standard population [20][21][22][23][24]. Directly standardized rates represent rates that would have existed if the population under study had the same age distribution as the "standard" population. Therefore, by being summary measures adjusted for differences in age-sex distributions, these rates afford more appropriate comparisons across the Portuguese ACES. ACES are grouping of essential PHC public services with administrative autonomy, consisting of several functional units, aiming to guarantee the provision of primary care to the population in Portugal, covering the entire country, taking into account the number of living people, the settlement structure, the rate of ageing and the accessibility to the referral hospital.

Relative Performance of the DM Prevention Quality Indicators across ACES
Preventable hospitalization rates observed in Portuguese ACES were summarized using the median and interquartile range per year analyzed since the PQI values were not normally distributed across ACES. The ACES (n = 55) were distributed across regions, 24 of them in the Norte region, nine in the Centro region, 15 in the Lisboa e Vale do Tejo region, four in the Alentejo and three in the Algarve (Figure 1). The median PQI value observed in 2017 was used as a cutoff point, one per PQI, to classify ACES in terms of their performance in each year: better performing ACES (PQI > cutoff); worse performing ACES (PQI < cutoff).
To compare the ACES preventable hospitalizations rates from 2016 to 2017, four possible scenarios were inferred: ACES that maintained throughout the biennium a preventable hospitalization rate lower than the 2017 median rate; ACES that decreased the preventable hospitalization rate relative to median 2017 rate; ACES that maintained throughout the biennium a preventable hospitalization rate higher than the 2017 median rate; and ACES that increased the preventable hospitalization rate relative to median 2017 rate. The former two scenarios are favorable, whereas the latter two scenarios are unfavorable. The most representative PQI scenarios within preventable hospitalizations and the composite PQI (which considers all hospitalizations due to diabetes complications) were analyzed and represented geographically in more detail. To compare the ACES preventable hospitalizations rates from 2016 to 2017, four possible scenarios were inferred: ACES that maintained throughout the biennium a preventable hospitalization rate lower than the 2017 median rate; ACES that decreased the preventable hospitalization rate relative to median 2017 rate; ACES that maintained throughout the biennium a preventable hospitalization rate higher than the 2017 median rate; and ACES that increased the preventable hospitalization rate relative to median 2017 rate. The former two scenarios are favorable, whereas the latter two scenarios are unfavorable. The most representative PQI scenarios within preventable hospitalizations and the composite PQI (which considers all hospitalizations due to diabetes complications) were analyzed and represented geographically in more detail.

Geographic Distribution of Relative Performance across ACES in Portugal
For a better perception and discussion of the results, in 2017 the Norte, Centro and Lisboa and Vale do Tejo regions were the most populous regions in the country exhibiting the highest number of hospital beds per capita (more than 310 beds/100,000 inhabitants). In contrast, Alentejo and Algarve had considerably lower population sizes and beds per capita (217 and 256 beds/100,000 inhabitants, respectively). However, Alentejo had a population density that is 3 times smaller than the next smallest population density region (Centro) and 11 times smaller than the highest population density region (Lisboa e Vale do Tejo). Unemployment rates ranged from 5.9% in Centro to 9.3% in Norte, and retention and dropout rates in primary education are smaller in Norte (9.3%), Centro (4.8%) and LVT (6.5%) when compared to Alentejo (7.1%) and Algarve (7.3%) (Table 1) [21,22].

Geographic Distribution of Relative Performance across ACES in Portugal
For a better perception and discussion of the results, in 2017 the Norte, Centro and Lisboa and Vale do Tejo regions were the most populous regions in the country exhibiting the highest number of hospital beds per capita (more than 310 beds/100,000 inhabitants). In contrast, Alentejo and Algarve had considerably lower population sizes and beds per capita (217 and 256 beds/100,000 inhabitants, respectively). However, Alentejo had a population density that is 3 times smaller than the next smallest population density region (Centro) and 11 times smaller than the highest population density region (Lisboa e Vale do Tejo). Unemployment rates ranged from 5.9% in Centro to 9.3% in Norte, and retention and dropout rates in primary education are smaller in Norte (9.3%), Centro (4.8%) and LVT (6.5%) when compared to Alentejo (7.1%) and Algarve (7.3%) ( Table 1) [21,22]. In order to assess the distribution of these preventable hospitalization rates for DM, an analysis was performed to understand the complications of diabetes PQI with greater representativeness in the composite and the results of the composite indicator itself (PQI 93) between ACES and the health administrative areas. A spatial representation of the trend scenarios for the 2016-2017 biennium of the PQI 03 and PQI93 results was made through the ArcGIS 10.5.1 software. This Geographic Information System (GIS) software is maintained by the global company "Environmental Systems Research Institute" (Esri) and allows to perform spatial analysis (Esri, Sacramento, CA, USA). It can be used as a support decision system allowing the visualization of information resulted from a large volume of data, in a multi-scale perspective [23][24][25][26][27]. It thus assists in the identification and direction of the problem solving and improvement of public policies, including public health [17,18,23].
The shapefile is representing the geographical influence area of each ACES and obtained from the National Health System (in WGS 84 (World Geodetic System 1984, EPSG:4326). The projected coordinate system used in the map was the ETRS89, PT TM06 European Terrestrial Reference System 1989 Portugal Transverse Mercator 2006, EPSG: 3763), currently the global reference system recommended by EUREF (European Reference Frame). Subsequently, the shapefile related information was combined with data from PQI 03 and PQI 93. The polygons on the map represent the ACES, which are classified according to the 2017 results scenario compared to the 2016 results, with the median of the PQI values' age-adjusted rate as the evaluation cut-off. The map was divided in 5 health administrative regions: Norte, Centro, Alentejo, Lisboa e Vale do Tejo and Algarve.

Results
The results of the age-sex-adjusted rates for diabetes prevention quality indicators described below were summarized in Table 2, and the data for the scenarios are presented in full in the tables of Appendix A.  In the unfavorable scenario, 13 out of 15 ACES which had a PQI higher than the median in both years also worsened their results. The two of the 15 ACES that were in this scenario but had a slight improvement were ACES Cova da Beira and ACES Pinhal Litoral. There were 12 ACES that raise more concern because had results below the 2017 median in the first year and were above in the second. Particularly, PQI rate in ACES Arco Ribeirinho increased from 10.5 to 32 hospitalizations/100,000 pop. in 2016 and 2017, respectively, the highest rate increase for this PQI among ACES.
With much lower rates and less representativeness within the group of indicators assessed, the median PQI 16-Lower-Extremity Amputation among Patients with Diabetes Admission Rate, in 2016, was 12.10 hospitalizations/100,000 pop. In 2017, it accounted for 202 admissions less than the previous year corresponding to a national median admission rate of 9.6 hospitalizations/100,000 pop. Despite this improvement, it was observed that in 2017 this PQI represented 16.51% of all DM preventable hospitalizations, while in 2016 it represented only 15.54%. In 2016, we observed only 1 ACES with results above the upper limit (quartile 3 + 1.5 × IQR), the ACES Baixo Alentejo (40.59 hospitalizations/100,000 pop.). In the following year, the only ACES above the upper limit was the ACES Estuário do Tejo (24.63 hospitalizations/100,000 pop.). As for the favorable trend scenarios for this indicator, 13 ACES remained below the national median of 2017 for both years evaluated. However, five of these 13 ACES showed an upward trend, despite still having results below the assessment value, with particular attention to ACES Médio Tejo, which reached 9.59 hospitalizations/100,000 pop. in 2017. The largest improvement in this scenario was ACES Pinhal Interior Norte, which rate more than halved from one year to the next (from 7.2 to 2.8 hospitalizations/100,000 pop.). Still, as a favorable scenario, 15 ACES that had rates above the 2017 national median, reduced their rates from 2016 to 2017. Of these 15 ACES the most significant improvement was observed in ACES Gerês/Cabreira, which went from 19.1 to 6.46 hospitalizations/100,000 pop. The unfavorable scenarios show 24 ACES with rates above the median in the two years, 13 of them with worse values in 2017 compared to 2016. The ACES posing the most critical concern in this group is the ACES Estuário do Tejo, which increased from 16.4 to 24.23 hospitalizations/100,000 pop. Moreover, the ACES that, despite being above the national median rate in the two years, improved its results was ACES Baixo Alentejo, decreasing from 40.59 to 20. For the biennium, the PQI rates interquartile ranges (IQR) are reduced, and the min-max values are closer from one year to other. The performance quadrants allow us to identify the ACES that are farthest from the defined cutoff, and their performance in the two years relative to the median 2017.

Geographic Distribution of Relative Performance across ACES in Portugal
PQI 03 deals with long-term complications of type 1 and type 2 diabetes-related to complications with renal, ophthalmic, neurological, peripheral, and other unspecified manifestations. As shown before, this PQI had the highest representativeness in the composite indicator (PQI 93). Figure 1 shows the spatial representation of the ACES referring to the trend scenario of the age-adjusted rate, for the PQI 03 and PQI 93.
It is essential to find out what the problem is and where it will be assessed. This study demonstrates that analyzing the composite in isolation can influence the interpretation of results. The map requires additional information so that it can target actions at the right time and to the right population. Figure 1A shows the results of the relative performance of ACES for the prevention quality indicator 03 (diabetes long term complications admission rate) for the biennium and 1-B for the prevention quality indicator 93 (diabetes complication composite rate).
As for PQI 03, the overall spatial representation showed a greater proportion of ACES (22 of 55, 49% of ACES) in the scenario characterized by a worst relative performance maintained throughout the two years, represented by 12 of the 24 ACES in the Norte Region, one of the nine ACES in the Centro Region, nine of the 15 ACES in the Lisboa e Vale do Tejo and one of the three ACES in the Algarve. The second most representative scenario within this PQI was that characterized by a relative performance improvement, with 21 of the 55 ACES (38.2%), eight of which belong to the Centro region, seven to the Norte region, five to Lisboa and Vale do Tejo and one to the Algarve. Only six ACES (five from the Norte region and one from the Lisboa e Vale do Tejo region) presented a good relative performance-maintained scenario. A single ACES presented a relative performance deteriorated (below the 2017 median in 2016 and above in 2017)-ACES Sotavento in the Algarve Region. The ACES with the highest preventable hospitalization rate in this PQI in 2017 was the ACES Arrábida, located in the region of Lisboa e Vale do Tejo (55 hospitalizations/100,000 inhabitants). ACES with a more substantial improvement from one year to the next (despite remaining above the median) was ACES Sintra also located in the Lisboa e Vale do Tejo region (from 56.95 to 26.25 hospitalizations/100,000 inhabitants) (Appendix A).
The results for PQI 93, a composite that include all preventable hospitalizations due to diabetes in the numerator (PQI 01, PQI 03, PQI 14 and PQI 16), show a higher proportion of ACES in the scenario characterized by a worst performance maintained throughout the two years ( Figure 1B). Twenty six of the 55 national ACES are found in this scenario (47.3% of the total), encompassing 11 of the 24 ACES in the Norte Region, two of the nine ACES in the Centro Region, 10 of the 15 ACES in the Lisboa e Vale do Tejo, three of the four ACES in Alentejo and one of the three ACES in the Algarve. The second most representative scenario within this PQI was that of good relative performance-maintained scenario, with 15 ACES present (nine in the Norte region, four in the Centro region, one in Lisboa e Vale do Tejo and one in the Algarve). Twelve ACES were on the scenario of relative performance improved (three Norte, three Centro, four Lisboa and Vale do Tejo, one in Alentejo and one in Algarve). Only one ACES's relative performance deteriorated with respect to preventable hospitalizations due to all complications of diabetes (ACES Santo Tirso/Trofa, in the North) which went from 56.43 to 67.73 hospitalizations/100,000 inhabitants). The ACES with the highest preventable hospitalization rate in this PQI in 2017 was the ACES Estuário do Tejo, located in the region of Lisboa e Vale do Tejo (120.6 hospitalizations/100,000 inhabitants). ACES with a more substantial improvement from one year to the next (despite remaining above the 2017 median) was ACES Baixo Alentejo, located in the Alentejo region (from 198.52 to 89.95 hospitalizations/100,000 inhabitants).
For the analyzed period, ACES Sotavento in the Algarve region, was the only ACES that worsened within its scenario, as it already had results above the national rate median and deteriorated further, that is, from 20.78 in 2016 to 28.43/100,000 inhabitants in 2017. Interestingly, the map shows us that, for this same region, the results made up in PQI 93 ( Figure 1B) are better performing. The same dynamic did not happen in the northern region, where ACES Santo Tirso/Trofa, presented to PQI 03 maintained a poor performance (above the median in the two years), and still worsened within its scenario (from 30.18 to 45.89/100,000 inhabitants). For PQI 93, there is a more significant contribution of PQI 03 in the burden of preventable DM hospitalizations. The darkest colour represents this ACES in Figure 1B.
In the Algarve region, we see a similar pattern of classification for two ACES, Barlavento and Central Algarve. The first maintained a poor performance for both indicators while ACES Sotavento, a relative performance deteriorated in PQI 03. However, for PQI 93 ACES Sotavento is in the category of good maintained performance. This result occurs since ACES Sotavento in PQI 01, and 14 had a good maintained performance with reduced rates from one year to the next. In contrast, PQI 03 and 16 had their performance deteriorated, being below the median in 2016 but above in 2017 (PQI03 = 28.43 and PQI16 = 13.82/100,000 inhabitants). Of the total of 439,617 inhabitants, 21.87% are elderly.
The Alentejo region is the region with the highest proportion of older adults patients who generally have multiple chronic conditions and are therefore more susceptible to long-term complications. It is, therefore, more targeted and personalized approaches to the chronic complications of DM that generate preventable hospitalizations (PQI03). In this region, however, there was an ACES that had its composite rate improved, ACES São Mamede, which went from 100.80 to 48.20/100,000 inhabitants (reduction of 47.8%). This region had this result in the composite as it managed to improve its performance in all indicators of complications (except PQI 03, where it remained slightly above the median in the second year, 26.35/100,000 inhabitants. As for the Centro region, eight of the nine ACES had improved results for PQI 03, and ACES Guarda was the only one that maintained a poor result for the two years. ACES Guarda, despite maintaining a result above the median in the two years, showed improvement in its result from one year to the next (from 32.21 to 28.66). This region is the third most populous in the country and has the third-highest proportion of elderly in the regions (25.41). In the analysis of PQI 93, we see a lighter map, which represents that four out of 10 ACES had good results maintained, emphasizing that ACES Guarda could still improve its results especially in PQI 03. An ACES in PQI 93 in the region maintained poor results ACES Cova da Beira maintained a poor result, impacted not only by PQI 03 but also by PQI 14 and 16. Despite having responded to an improvement in their values from one year to the next, they maintained a poor performance in the biennium compared to national results.
The Lisboa e Vale do Tejo Region, the most populous region of the country, presented in PQI 03 a single ACES with good maintained performance, and a result also demonstrated in PQI 93 (ACES Medio Tejo). Five ACES had their results improved in PQI 03, with reduction of rates in all of them, in particular, ACES Lezíria, which went from 40.26 to 16.11/100,000 inhabitants). Although nine ACES maintained worse maintained results, all of them had reduced rates from one year to the other except for ACES Arrábida, the highest national rate 55/100,000 inhabitants.
In the North region, specifically further south of the region, we observed 12 ACES that maintained the results of their rates from one year to the other concerning PQI 03. However, of those seven ACES maintained rates below the median in the two years, but signalled a slight increase rate from one year to the next (ACES Alto Minho, Ave/Famalicão, Aveiro Norte, Baixo Tâmega, Feira e Arouca, Gaia, and Vale do Sousa Sul). ACES Santo Tirso/Trofa is the darkest area, previously described.Therefore, despite presenting better results in the year of 2016 among the groups of indicators assessed, the individual scenarios for ACES have individual trends that requires more considerable attention from services to diabetic patients (Appendix A).

Discussion
To the best of our knowledge, this is the first study that presents the distribution of preventable hospitalization rates for diabetes complications in Portuguese territory. It is also the first one to analyze these results through performance scenarios of trends.
Portugal has endeavored, through its national diabetes prevention and control program, to reduce not only rates of hospitalization due to complications but also mortality due to DM. Therefore, Portuguese PHC has been complying with the commitments signed with the World Health Organization and reaffirmed in the Declaration of St. Vicent [28] of which Portugal participated. Expertise in Primary care management of diabetes is naturally uneven but to identify those practices that need help to improve care necessitates methods to identify these sites.
The PQI values for DM established by AHRQ give us a direction for strengthening primary health care the main responsible for the management of treatment and follow-up of these patients. The absolute number of preventable hospitalizations for all diabetes complications in Portugal has decreased by 1726 hospitalizations (20%), thus passing from the rate of 79 to 65 hospitalizations/100,000 inhabitants. The PQI 03, which deals with long-term complications, was the most representative of the total number of preventable DM admissions, corroborating the natural history of the disease and reflecting control efforts at the primary level [29][30][31][32]. These long-term complications (despite being the leading causes of hospitalizations in the two years) had a significant reduction in the absolute number of preventable hospitalizations-from 4012 hospitalizations to 2639 (reduction of 34%). Despite the improvement in results for PQI 03, 20 of 48 ACES that were above the median of 2017 in 2016, achieved better results the following year, and for the composite of preventable diabetes hospitalizations (PQI 93) only 11 out 39, revealing the need for further studies and actions to improve the diabetic quality of life.
We recognize that the present study may have some limitations. First, it is important to remember that our study does not intend to assess barriers on access to health care services and predisposing factors of preventable hospitalizations for diabetes mellitus, but only to map an overview of the performance of primary health care regarding hospitalizations for complications of an ambulatory care sensitive condition in the biennium in question, in order to identify the groups of health centers that demand greater attention from health managers and professionals, with intensification of health promotion, disease monitoring actions, and future micro-analyzes at the level of the units belonging to these groupings.
Secondly, we were unable to consider many potential confounding factors. The high heterogeneity found in the rates of hospitalization for complications from DM may suggest differences in critical areas in the quality of primary health care. The indicators presented may be associated with some factors that are external to the health system, such as socioeconomic level, environmental conditions. lack of patient compliance with treatment, data not available for the studied aggregation of ACES [33][34][35][36][37][38][39][40][41]. ACES were used as a unit of analysis because in Portugal, the district health network is administered independently in accordance with national policy. Comparisons with other countries should be treated with great caution, as in previous studies the selected geographical units are varied [12,35,36,41]. Barriers to access care for diabetic patients, such as the geographical barrier or the absence or low availability of caregivers or family members to conduct to the consultations, especially older patients, even the existence of family health units can also be factors that can interfere with the results presented. Again, these data were also not available for analysis. Although the results presented are engaging in the sense of signaling the greatest need for intensification of disease prevention and control actions, we must exercise caution in interpreting the data as an absolute truth since the analysis period reported only to two years. Therefore, the methodology of this study must be applied later in a more extended period, in a prospective study.
This article brings added value in terms of knowledge about trends in the results of primary health care actions in the territory, concerning the unfavorable outcome (avoidable hospitalizations). It provides a spatial demonstration of the results, contributing to identify locations where public policies should compose more consistent actions of continuous quality improvement, like in access, follow up, continuity, treatment, health education, among others. Besides the analysis can be replicated, it can also be improved with more details at the local level, observing factors that can contribute to the improvement of results, such as associations with quality indicators of primary health care, socioeconomic and human resources allocated for treatment and monitoring of these patients. Furthermore, it may be the starting point for the analysis of other PQI rates such as the Chronic Obstructive Pulmonary Disease, Hypertension Admission Rate, among others. Several studies have been conducted and published in the literature regarding PQI, in many approaches, from the simplest to the most complex [32,[47][48][49][50][51][52], that are possible allied methodologies for a more in depth comparative analysis, which can be reinforced and based on this method of identifying areas through their respective quadrants of performance scenarios, still to be performed by our group in Portugal. The analysis of quality indicators related to diabetes in PHC and their association with the variability of the results of preventable hospitalizations, in addition to the efficiency analysis of the resources applied in PHC, are ways to deepen this analysis, either individually in each ACES as per regions.

Conclusions
Our study identified a significant rate reduction from one year to the next, and PQI 03 (long-term diabetes complications) as the most representative from the total preventable hospitalizations in the two years. Despite the performance improvement in, we notice that many ACES at the national level remain with rates above the 2017 national median, both for PQI 03 and for the composite of preventable diabetes hospitalizations, revealing the need for further studies and actions to improve the diabetic quality of life. The benefit of the quadrant scenarios through spatial analysis method to identify low-performance primary care facilities has proved to be practical and a suitable visual method for signaling ACES with performance standards variation for diabetes complications indicators and their impact on the results of the composite PQI across ACES.
Author Contributions: The study conception was performed by A.R. and A.F. All authors contributed to the study design. Material preparation, data collection and analysis were performed by A.R., L.D., M.L., J.S., P.S. and A.F., A.R. wrote the first draft of the manuscript and all authors commented on previous versions of the manuscript. L.D. was responsible for processing the data in the GIS software for the maps. All authors read and approved the final manuscript. The authors are responsible for the correctness of the statements provided.

Funding:
The authors disclosed receipt of the following financial support for the research, authorship, and publication of this article: This work was financed by FEDER-Fundo Europeu de Desenvolvimento Regional funds through the COMPETE 2020-Operacional Programme for Competitiveness and Internationalisation (POCI), and by Portuguese funds through FCT-Fundação para a Ciência e a Tecnologia in the framework of the project POCI-01-0145-FEDER-030766 ("1st. IndiQare-Quality indicators in primary health care: validation and implementation of quality indicators as an assessment and comparison tool").

Acknowledgments:
We like to thank the Central Administration of the Portuguese Health System (ACSS) for providing the administrative data on hospitalization, and to MSc Hugo Manuel Ribeiro Teixeira from the University of Porto for his support for the final spatial representation using the ArcGIS software.

Conflicts of Interest:
The authors declare no conflict of interest.