Many variables may play a role in patient mobility: distance, type of hospital/unit, expected outcome, costs, medical staff acquaintances, advice from physicians or friends, previous experiences and waiting lists [3
]. Our research was based on the assumption that patient mobility is partly influenced by active choice.
It is easy to determine that patient mobility is influenced by active choice on the part of a patient, possibly guided by his/her GP, when the hospital is distant from the patient’s home. For example, studies have highlighted distances of more than 1000 km covered by patients living in southern Italy hospitalised in the north [5
]. Another form of mobility, termed “physiological” [5
], is influenced less by active choice than by reasons unrelated to the quality of care. Such a reason is geographical position: in border-areas it may be rational for patients to access cross-border care if this implies shorter distances [2
]. In this case, crossing the border of the LHA may express a preference dictated by easy access rather than by better care. In some cases patients move without making any rational choice [4
], whereas in others, when patient condition is at least moderately severe or complex [5
], patients are driven by their own active choices or by those of their referring GPs. In the present study, the Gandy nomogram, trends and APR-DRG severity stratification described general aspects of patient mobility and also identified some unexpected situations.
The main findings of the study were as follows:
i) Movement of the points from left to right in non stratified nomograms showed that all HAs were losing capacity to attract patients and gaining capacity to avoid escapes. The trends showed that HA1 and HA2 succeeded in regaining lost patients, whereas HA3, despite a positive trend, was slow in doing so. Patient trust in health services may have important repercussions on future trends. Retaining/attracting patients takes time and persistence, while loss and distrust of patients have major consequences.
ii) Aspects embedded in the phenomenon of patient mobility emerged with severity stratification. Trend analysis highlighted that while all HAs showed similar trends of unstratified mobility at the zero-minor severity level, at the moderate level HA1 maintained its position (top upper right quadrant), HA2 had more escapes than attractions (lower part of upper right quadrant) and HA3 descended into the sector where escapes are more numerous than attractions and the return of residents. Stratifying at the moderate severity level, HA1 and HA2 maintained or improved their position over the years, especially in recovering escapees. This aspect is particularly important because it shows that patients could tend to remain in the same HA even when their condition is moderately severe, probably demonstrating trust in the health services where they live. Health Area 3 had all its points in the lower part of the upper right quadrant of the nomogram and in the quadrant where escapes are more numerous than attractions and recovery of escapees. The numbers were few and therefore percentages more likely to vary in position, and no trend was detected. The broad cloud of points may indicate a will to improve position, though it is not easy to return to the upper quadrants.
While extreme severity should limit patient mobility because of the objective difficulty of moving critical patients (points should be in the middle of the upper right quadrant), when patients do indeed move of their own initiative, it is very informative. In seeking medical care far from home and family, in a different environment, and paying all associated expenses [2
], the patient votes twice, expressing distrust of the service available locally and belief in obtaining better treatment elsewhere. For the hospital, this patient’s choice is a double failure, indicating loss of patient trust and a lost opportunity to treat the patient, with negative economic consequences as well. The stratified nomograms provided a description that was not possible with the unstratified data. Differences in hospital selection exist and are influenced by disease severity. Moreover, trend analysis showed important changes underway: patients have modified their preferences for different hospitals. Whether these changes are influenced by the real or perceived quality of care is relevant to what measures should be taken. If there is an objective lack of service/care, specific improvements are necessary to limit escapes. If the changes are only driven by patient perception, managers should improve communication with the public and general practitioners [41
]. Involvement of the latter is particularly important in Italy where GPs are patients’ first contact with the health system. Since general practitioners’ advice may influence patient choice [20
] it is necessary that GPs be informed and kept up to date on LHU services. Health planners should ensure equal distribution and access of services to patients: the escape of patients could highlight a problem in achieving this aim.
iii) Comparing unstratified with severity-stratified data, it emerged that it was easier to maintain a position in the nomogram if the points were in upper part of the top right quadrant than in the lower part. The points of HA1 and HA2 generally maintained their positions, irrespective of APR-DRG stratification, whereas HA3 did not. An important difference revealed by stratified analysis was that as severity increased, the possibility of retaining patients decreased. Patients move to other areas when they feel that local health services/doctors are unable to properly care for them.
This descriptive study was conducted to acquire knowledge in an increasingly important field of public health. Stratification is always needed when comparisons are made, as we found in our study. Patient mobility also calls for severity stratification, although in the present case it was already partially selected and regarded a specific disease group. The APR-DRG system seems to identify some critical aspects, highlighted by the stratifications, and to provide useful details for investigating patient healthcare choices at different severity levels.
The differences identified studying cardiac surgery admissions are likely to exist for other diseases as well, but further studies are necessary. Other risk adjustment tools, such as the Charlson [42
] or the Elixhauser Comorbidity Index [44
], which use hospital discharge records as a basis for determining severity scores, could be used to verify our findings and to conduct sensitivity analysis [45
]. For example, the APR-DRG produced very few cases at the extreme (severest) level (3), which prompted us to discard this small sample.
On one hand, risk adjustment tools do not cover all the possible aspects of confounders: they are not definitive in assessing borderline differences. Although widely used and improving in time, hospital discharge records only contain what are considered to be major health conditions. On the other hand, risk adjustment tools can provide useful indications if differences are large, as we have seen with stratification and trend.
The study has certain limits: i) It does not quantify “physiological mobility”, that is, movements from areas near Regional/HA borders, unrelated to improvements in heath care quality [5
], but due to closeness/easy access to hospitals in other areas (Tuscan HAs and neighbouring Regions of Italy). Physiological mobility is more evident in neighbouring areas. ii) Other factors, not detected by this study, could have influenced patient mobility. For example, hospitals may influence the referral pattern of GPs. In our study this possibility should be limited, because the hospitals/HAs had similar characteristics and offered similar services. It is more likely to occur when differences exist. The fact remains that GPs may refer patients to one hospital or another for various reasons, such as: locality and convenience, familiarity with hospital, good overall service, sub-specialist care, clinical need of patient, waiting time for appointment, good communication at hospital or a known consultant [47
]. iii) It may attribute patient mobility to citizens officially resident in areas different from those in which they actually live [48
]. Such mobility is not easy to detect and is difficult to measure; it may also be influenced by several factors. A previous attempt to estimate this portion of mobility [49
] assumed that it could be equivalent to: i) mobility for diseases without complications, treatable in any hospital and ii) emergency admissions where patients cannot choose [38