In recent years, a focus on population health and addressing health inequalities has become a major public health research priority area. Central to this endeavour is the use of health indicators, which serve as tools in assessing and monitoring the health statuses of populations. Health indicators include a mixture of statistics, measures health service utilisation, and surveys which often include reports on service utilisation and quality of life measures [1–3].

A 2021 systematic review reported the use of 691 health indicators covering 120 topics, most commonly using measures such as life expectancy, infant mortality, prevalence of obesity/overweight, mortality rates, behavioural issues, the presence of health conditions and healthcare utilisation [1].

Healthcare utilisation, a key measurable component of population health, plays a critical role in understanding health outcomes and access to healthcare services. The patterns of healthcare utilisation within a population not only reflect the distribution of health needs but can also shed light on inequalities in access to care [4, 5].

Health indicators also feature in broader societal developments that influence policy development. For example, good health and well-being, gender equality and reduced inequalities are three of the United Nations Sustainable Development Goals [6].

However, many health indicators suffer from a lack of available data which can inform policy development. Desirable qualities include ease of interpretation, sensitivity to policy changes, timeliness, historic time series, duality in measuring physical and mental health conditions, ability to be updated regularly, and disagreeability by population subgroups [2].

With these considerations in mind, and most notably the availability of population-wide routine data sources to fill gaps in the literature, we conceived a parsimonious, reproducible, temporally flexible, population-wide indicator which could be derived from routine data sources and would be easily understood by policymakers and the general public. The Desirable Health Indicator (DHI) was conceived at a European Information for Action (InfAct) meeting in 2014 on population health indicators but not fully implemented until 2024 by the Administrative Data Research Wales group [7, 8]. The concept arose during discussions concerning novel population health indicators that would reflect a positive view of health in line with the United Nations Sustainable Development Goals [6]. This led to scenario planning in which health was so well protected that few people became ill, and for those that did experience illness, primary care provided efficient and effective care. In such a scenario there would be little need for hospital or emergency care as these were viewed as system failures.

One of the rationales behind the indicator was to keep it as simple as possible in terms of data requirements in order to allow for replication in many different settings, including in those without general practice data but with access to dispensing data. General practice data is needed for measurement of the prevalence of many chronic conditions[9]. We did consider prescriptions for individual morbidities but did not include these as many medications have multiple uses. It is extremely difficult, if not impossible, to develop a one to one relationship between the medication and the illness, for example beta-blockers which can be used to treat angina, arrhythmias, hypertension and a variety of non-cardiac condition.

Many treatments for chronic diseases are at least partially effective and allow individuals to live lives without the need for emergency health care interventions. We consider that such people should also be candidates for meeting the DHI.

The DHI concept operationalised in this paper is based on the idea of what a person would wish for in the forthcoming year in terms of health. It comprises six sub-indicators including living for the entire year (i.e. survival), absence of hospital activity (inpatient and emergency department attendances), and absence of prescriptions for common conditions that people would wish to avoid (namely medication for infections, analgesics, or mental health conditions). The aim of this paper is to describe the creation of the DHI and describe its distribution in the population of Wales, UK. Such an indicator could be implemented in many countries as data linkage capabilities grow [10].

Figure 1 shows the number of people registered with Welsh GPs at the beginning of each year and the number who were in the 86% of GPs that supply data to SAIL.

Figure 2 shows the percentage of the population that met the DHI each year of the study period. Results for 2015 to 2019 are similar, with a marginal but gradual increase in the percentage of individuals meeting the DHI over this period. In 2015, 39.6% of the total population met the DHI, increasing slightly to 41.9% by 2019. The 2020 and 2021 in the figures were 48.6% and 46.2% respectively, reducing to 43.1% in 2022.

Given the large number of results and small differences between years we focused on the year 2022 for the remainder of the analysis. Overall, 2,677,829 people were included in the analysis, with 1,154,630 (43.1%) identified as meeting the Desirable Health Indicator with 49.2% of males met the indicator compared to 37.1% of females. The highest percentage by age group was the 10–19 year-olds at 58.5% meeting the DHI, compared with 12.8% in the oldest age group (90+) (Table 1). Nearly half of the population who live in the least deprived (most affluent) areas of Wales met the indicator (47.2%), compared to 39.4% of individuals who live in the most deprived areas (Table 1).

Table 2 reports the univariate and multivariate logistic regression analysis results to measure the odds of meeting the DHI by the main effects: age group, sex, and deprivation quintiles. Being male (aOR=1.62, 95%CI = 1.61–1.63), aged 10-19 years (aOR = 1.69, 95%CI = 1.68–1.71), and living in the least deprived areas of Wales (aOR=1.31, 95%CI = 1.30–1.32) had the highest odds of meeting the DHI. Conversely, the oldest age group Wales had the lowest odds of meeting the DHI (aOR=0.18, 95%CI=0.17–0.19) and when looking at deprivation, those living in the most deprived areas of Wales had the lowest odds of meeting the DHI (aOR = 0.79, 95%CI = 0.78–0.80). S26 contains the multivariate logistic regression analysis results to measure the odds of meeting the desirable health indicator by the main effects with interactions.

The percentage of individuals meeting all six sub-domains of the DHI decreased with age and with increasing socioeconomic deprivation (Figure 3). Across all age categories, a higher percentage of males met the indicator, except for 0–9 year-olds (Figure 3). Similarly, across all age categories and both sexes, there was a higher percentage meeting the DHI in the least deprived areas. The demographic group with the highest percentage of individuals meeting the DHI (64.9%) were 20–29 year old males from the least deprived areas of Wales (Figure 3). Conversely, the demographic group with the lowest DHI percentage (9.02%) were the 90+ year old females living in the second most deprived areas of Wales (Figure 3).

In total, 1.15% of the 2022 cohort died, 12.5% had a hospital admission, 19% attended the emergency department, 29.5% were prescribed drugs used in infections, 22.8% were prescribed analgesics and 20.2% were prescribed drugs for mental health (Table 3). Across all socioeconomic groups prescriptions for drugs used in infections were the most common reason for not meeting the DHI (27.2-31.4%) (Table 3). For those aged below 40 years, prescriptions for drugs used in infections were the main cause, while prescriptions of analgesics were the most predominant among those aged 50-90+ (Table 3). In the age group 40-49 years, mental health prescriptions stood out as the most frequent reason for not meeting the DHI (48.3%). Prescriptions for drugs used in infections had the highest prevalence when looking at both sexes (Table 3).

Females exhibited a higher percentage of healthcare service utilisation compared to males (Figure 4). The percentage of people being admitted to hospital, being prescribed analgesics, and dying increased substantially with age (Figure 4b, 4e, 4a). Additionally, individuals from more deprived backgrounds attended emergency departments and received prescriptions for drugs used in infections, analgesics, and mental health more than those from the least deprived areas (Figure 4c, 4d, 4e, 4f). Results and figures for all other years (2015-2021) can be found in (S2-S25).

This study aimed to address the challenges of creating and effectively communicating population health metrics by introducing an easily interpretable indicator, the Desirable Health Indicator (DHI).

The study followed the health records of 2.6+million people over an 8-year period. Trends in DHI percentage were stable between 2015 and 2019 (39.6%-41.9%) with a notable increase across 2020 and 2021 during the height of the COVID-19 pandemic (48.6% and 46.2%), with a decrease in 2022 (43.1%) (Figure 2), consistent with published UK and international studies on the negative impact of the COVID-19 pandemic on healthcare service use and provision [17–20].

Focussing on results for 2022, less than half of the population (43.1%) met the criteria for the DHI (Table 1). In total, 1.15% of the total study population died in 2022, 12.5% had a hospital admission, 19% attended the emergency department, 29.5% were prescribed drugs used in infections, 22.8% were prescribed analgesics, and 20.2% were prescribed drugs for mental health (Table 3). When examining the differences demographic groups, males are more likely to meet the DHI (aOR = 1.62, 95%CI = 1.61-163) compared to females (Table 2). Across the sub-indicators, females are more likely to interact with health services (Figure 4, Table 3), and particularly to be prescribed treatment for mental health, painful conditions and infections. This is consistent with existing literature showing the prevalence or incidence of these conditions to be more common in this group [21–23].

Another potential explanation is the influence gender has on health seeking behaviour with females having consistently higher consultation rates [24, 25].

Failure to meet the DHI was substantially higher for people living in the most deprived areas (aOR = 0.79, 95%CI = 0.78-0.80) compared to those living in the least deprived areas (aOR=1.31, 95%CI = 1.30-1.32) (Figure 2 and Table 2). When combining age, sex, and area-based deprivation to examine outcomes in specific demographic groups, 20-29 year old males living in the least deprived area of Wales had the highest percentage of individuals who met the DHI (64.9%), compared to 9.02% of 90+ year old females living in the second most deprived areas who had the lowest percentage (Figure 2).

Individuals living in the least deprived areas consistently had a higher percentage of individuals who met the DHI (47.2%), contrasting with 39.4% in the most deprived fifth (Table 1, Figure 2). For the most deprived groups, there are higher percentages of prescriptions for analgesics, mental health drugs, and infections, a higher use of emergency departments but very similar percentages with hospital admissions (Table 3, Figure 4). A meta-analysis of 45 studies from 12 mostly western high-income countries, reported that chronic pain prevalence is found to be substantially higher in more deprived communities and individuals [26]. Likewise, existing literature shows much greater prevalence of mental health conditions in deprived areas and individuals [27, 28]. Our findings are consistent with the analysis of emergency department attendances by deprivation in England conducted by the Office of National Statistics [29]. The finding that hospital admissions were similar across deprivation fifths contrasts with some of the literature which reports higher rates in most deprived groups in England [30]. It should be noted that deprivation fifths are based on the Welsh Index of Multiple Deprivation and will have different cut-offs to the distribution of the Index of Multiple Deprivation in England [31].

The DHI was conceived as a parsimonious indicator that would be relatively easy to compute. It covers the range of conditions identified by the SIPHER (System science in Public Health and health Economics Research) desirable qualities of an indicator, which include easy interpretation, sensitivity to policy changes, timeliness, historic series, duality in measuring physical and mental health conditions, ability to be updated regularly, and disaggregation by population subgroups [2]. There are many specific conditions that could be included but the DHI includes metrics on mental and physical health and infection and hence covers many bases.

The concept behind the DHI was well received by policymakers, representatives of public health agencies across Europe, and members of the public. They reported that it gave them easy to understand and novel insights into population health with a desire to see implementation in multiple countries and suggested further work to add a deeper understanding of the components which lead to failure to meet the DHI.

Members of the public would like to see further developments and analysis in relation to: understanding and explaining the inequalities by deprivation group; expanding analyses to more groups of people such as ethnic minorities and immunocompromised people; adding more layers of context, such as distinguishing reasons for attending the ED (injury vs medical problems); and conducting analysis by groups receiving one prescription versus multiple, and by type of painkiller e.g. paracetamol, anti-inflammatories and narcotics.

Following literature review and discussion with public health experts in PHIRI, the DHI appears to be a novel indicator. Most research on health service utilisation and health indicators tends to use metrics, such as age standardised rates per 100,000 for conditions, disability adjusted life years, life expectancy, and daily defined doses per 100,000 persons or consultations, for example the Healthy Belgium website [32]. These are helpful for comparisons between regions and across time and for public health experts, but do not immediately resonate with the public or many policymakers as the constructs are difficult to grasp for non-scientific audiences. One of the key strengths of the DHI is its simplicity as a reproducible, population-wide, and easily understood health indicator.

The individual components of the DHI or their proxies have been well studied in the literature but not assembled as an indicator. There are many studies and websites that compare mortality, hospital admissions, emergency department attendances and measurements of prescribing for antibiotics, mental health conditions and pain but report using different metrics, such as daily defined doses (DDD) per 1000 persons per day for antibiotic prescribing or per consultation. The DHI sub-indicator is on prescribing for all infections and hence wider than antibiotics prescribing, whereas DDD metrics include repeat prescriptions per individual and differing defined doses per drug, limiting comparability.

Among the strengths of the study is the ability to link multiple anonymised demographics, longitudinal primary and secondary healthcare, and mortality data for the population of Wales, UK through the SAIL Databank, creating very large, population representative studies, with outcomes recorded in different datasets [11, 12].

Whilst our study has provided important findings, several limitations need to be considered. The DHI was designed to be parsimonious and easy to implement for jurisdictions with limited data assets. It could be criticised and in that it does not contain a comprehensive list of health conditions but deciding on which health conditions to include is not an easy task.

General practice data is needed for measurement of the prevalence of many chronic conditions [9]. We did consider prescriptions for individual morbidities but did not include these as many medications have multiple uses. It is extremely difficult, if not impossible, to develop a one to one relationship between the medication and the illness, for example beta-blockers which can be used to treat angina, arrhythmias, hypertension and a variety of non-cardiac condition.

Many treatments for chronic diseases are at least partially effective and allow individuals to live lives without the need for emergency health care interventions. We consider that such people should also be candidates for meeting the DHI.

It should be noted that the DHI resonated well with policymakers and members of the general public. The DHI is not a direct measure of health, it measures aspects of health and indications of likely absence of conditions that are usually treated with prescriptions. It will also be influenced by issues such as personal resilience, health seeking behaviours, how clinicians manage conditions, and patterns of health service provision. Additionally, this indicator utilises data held in the SAIL Databank which contains 86% of Welsh GP data and so whilst this is intended as a population-wide indicator, it does not yet cover the complete population. However, analysis has shown small differences between participating and non-participating practice populations that do not bias the findings [33]. A further limitation is that our analysis of prescribing is limited to general practice issued scripts as electronic hospital prescribing has not yet been rolled out nationally. Some 70% of antibiotics are prescribed by GPs in the UK. However, many prescribed antibiotics in hospital are likely to have also prescriptions from GPs, limiting this bias [34]. Finally, this indicator has been developed on anonymised NHS data, SAIL does not contain private health records and so any private prescriptions or elective operations in private hospitals are not included. Private healthcare use in lower in Wales due to the free to use NHS system [35].

Future work will include considering the recommendations of the members of the public listed above. Permission has been given to incorporate Census 2021 data into the DHI to include outputs on demographic and protected characteristics such as ethnicity, gender identity, and disability as well as socioeconomic status. We are also seeking access to a dispensing database that covers 100% of the population and to datasets with measures of income and benefits to conduct analysis of household and individual level socioeconomic status. Additionally, with other population-wide linked data systems available, we would like to compare the Welsh results with other nations or populations. The method has been shared with the European PHIRI group to aid dissemination and replication. We will also use longitudinal data to develop predictive modes forecasting the Desirable Health Indicator and its components in subsequent years.

Implementation of the DHI across various settings and jurisdictions, providing accessible and comparable data, will offer an interesting insight into variation by healthcare systems and opportunities to gain insight into underlying reasons. Being able to calculate the DHI by various subgroups of the population highlights disparities in outcomes between patient groups and helps identify groups for improved policies and intervention. For example, the high proportion of the public being treated with mental health drugs is not an indication that prescribing for mental health problems should stop but an indication that there are serious problems with mental health in a population that have not been adequately addressed by current policies.

In conclusion, this study has made significant strides in bridging the gap between population metrics and public understanding by introducing the DHI as an innovative, accessible, and comparable indicator of population health and healthcare utilisation. We demonstrate that this indicator provides clear insights into healthcare service utilisation patterns in Wales, specifically examining variations across socioeconomic status, sex, and age groups. Moving forward, informing policymakers and the public about healthcare utilisation patterns can facilitate evidence-based decision making and resource allocation. Research into the determinants of healthcare utilisation will help identify key factors driving disparities and inform targeted interventions. Furthermore, evaluating the effectiveness of embedded interventions aimed at reducing the undesirable use of healthcare systems can guide efforts to improve their efficiency and equity.