Predictors of a depression diagnosis were a high number of physician visits, female gender, age below 60, age above 80 and a low socioeconomic status.
Patients who received a diagnosis of depression used twice the amount of health care (e.g. physician visits and hospital days) during the five year period prior to diagnosis compared to the control group. A particularly strong increase in health care utilisation was seen the last year before diagnosis. These findings were supported with a high level of co-morbidity in terms of both RRCW and particular diagnoses, as for example musculoskeletal disorders, during the whole five-year period for patients with a depression diagnosis.
Co-morbidity was an important factor in order to explain the high level of resource use for patients with a depression disease. From many studies it is known that the musculoskeletal system is overrepresented in patients who later receive a depression diagnosis . The same is valid for the patients with problems from the cardio-vascular system . This fact was also confirmed in our study where 60% of the patients with a depression diagnosis also suffered from musculoskeletal disorders and 36% had a cardiovascular disease before getting the depression diagnosis.
Depressions are more common in midlife than in the elderly. However, depressive symptoms are more frequent among the oldest old explained by factors associated by aging such as co-morbidity . This is also seen in our data: odds ratios were highest in young ages and decreasing in the elderly and finally with a peak in the oldest old (but with lower odds ratios than in midlife). More adjustment for co-morbidity (i.e. physician visits) reduced the depression odds ratio in the older groups illustrating that co-morbidity is an important factor especially in the oldest old (Model 3 versus Model 1, Table 2).
Psychiatric diagnoses have been found to be underreported in registers, especially in primary care . Drug use, might be another important indicator of high health care utilisation and also a predictor of a future depression diagnosis. Hence, the use of drugs with a somatic indication has been found to be on a high level several years before treatment with antidepressants and prior to a psychiatric diagnosis .
In the study by Bingefors et al , only 45% of antidepressant-treated patients had a recorded psychiatric diagnosis. But when the medical records for these patients were further scrutinised 80% had depressive symptoms or a diagnosis of a depressive disorder. In our study it was shown that patients treated with antidepressants used more than twice as many health care resources as controls. Earlier studies of depression patients have shown a high concurrent use of health care resources, also for somatic complaints .
In our study almost 30% of the patients in the group with a depression diagnosis were treated with antidepressants during the year before diagnosis, with a majority of the patients between 45 and 60 years of age. The corresponding share of the control group was only 6%. Our finding that not all of the patients with a depression diagnosis receive antidepressants is in line with other studies .
Treatment with antidepressants increased to more than 80% of the patients the year after the depression diagnosis was given, compared with 7% in the control group.
This implies that 20% of the patients with a depression diagnosis were still not treated with an antidepressant drug. Some of these patients may have been treated with other alternatives such as physiotherapy or cognitive behavioural therapy.
The discrepancy between psychiatric diagnoses and the antidepressant drug use might be explained by doctors and patients behaviour. Health care providers may record the same condition with a variety of cods and it might be possible that a depressive health status was present before the depression diagnosis put in the medical records. A somatic diagnosis might have been used instead because psychiatric conditions might still be stigmatized by society and patients would perhaps rather see their primary physicians than psychiatric specialists.
There is a lack of population-based longitudinal studies examining the relationships co-morbidity, and health care utilisation. However, in a similar investigation compared to our study it was found that obstructive sleep apnea syndrome patients have symptoms, high health care utilisation and co-morbidities as far back as ten years before recognition of their disorder .
Strengths of this study were the use of different register databases and the linkage to other registers. The use of a personal identification number made it possible to follow individuals and there was total population coverage. The possibility to follow health care utilisation on an individual level during a long period of time even in primary care was a great advantage in this study. When using registers, sources of bias such as recall bias, response bias and selection bias could be kept at a minimum. An additional strength was the size of the study, with more than 2400 patients having a new diagnosis of depression. The study size results in a high power for the analysis of interactions and, since none of those were even close to significant, our results indicate that the stability of the associations between health care utilisation and later depression diagnosis over age groups, gender, and socio-economic status is high.
A weakness of using registers is the quality of data and the broad clinical variation, for instance variation in the definition of depression. This might create a bias and an uncertainty in the cohort of depression diagnosis.
The co-morbidity concept is complex and there is a lack of consensus about how to define and measure the concept. Other concepts are multi-morbidity, burden of disease and frailty, especially in the geriatric field . Often these concepts are multidimensional and more comprehensive compared to the measure RRCW we have used in the present study. However, the ACG instrument has been used and been found of a good validity in a Swedish setting  and it was easy to handle in our study. In this study the number of physician visits and RRCW were highly correlated, which is not so unexpected since RRCW is constructed to predict health care utilisation. One consequence was that physician visits and RRCW could not be included in a regression model simultaneously without a considerable loss in statistical power.
There are several factors associated with the heavy health care utilisation among patients with a depression diagnosis. We adjusted for well-known factors associated with high health care utilisation such as a low socioeconomic status and female gender . Female gender usually has a high use of health care resources and it is well-known that women have a high incidence and prevalence of depressive disorders . A weakness of study is that we only analyzed patient-related variables and not provider-related variables such as location of health care units (rural or urban), private or public care, remuneration system, etc. which might influence health care utilisation. However the Swedish health care system and the County Council of Östergötland was (in 2006) rather homogenous, geographically and economically, and was dominated by public providers and a health care personnel who was salaried.
Patients with a depressive diagnosis represent a large share of patients in health care, especially in primary care. GPs are nowadays confronted with an increasing multi morbidity of an aging population and they are under pressure of psychiatric hospitals which seek to integrate their patients in the local community. The organizational format is to attach different professionals to GP practices. A central obstacle in Swedish health care is that clinical services are the responsibility of county council while care for social needs is the responsibility of the local municipalities. Besides, there is no GP gate-keeper system in Sweden. However, a new reform: "choice of care" for citizens has been introduced within primary care where different actors, both public and private, provide services as long as certain quality criteria are met. Today primary care services are therefore provided by a mix of public and private actors which may influence clinical management and sometimes create delays in the health care processes. However, there are also recent organisational incentives to convince primary care, specialist care and social care to create "chains of care" for chronically ill persons and patients with multi morbidity.
In this study we found that early signs of a depression diagnoses were a high number of physician visits, female gender, age below 60, age above 80 and a low socioeconomic status. The incentives to create chains of care might therefore include and integrate our findings in order to prevent delays in the clinical patient management. Early diagnosis and treatment of patient with depressive disorders would improve the health situation of these patients and may reduce the financial burden of health care providers and taxpayers. The effectiveness of pharmacological treatment of depressive disorders is generally accepted, and most patients who are treated experience a strong improvement in their quality of life and are able to take part in a daily working life.
To conclude, the high level of co-morbidity during the whole five-year period for patients with a new depression diagnosis had a heavy impact on health care utilisation. Caution should be exercised in the GP surgery when meeting a young or very old woman with a low socioeconomic status and many physician visits within a short time-period. Early detection and adequate treatment of these patients may reduce high health care utilisation and improve the quality of life for this group of patients.