Hospital Transfer Between Primary and Secondary Metabolic Bariatric Surgery in The Netherlands: A Cross-sectional Multi-party Computation Analysis of Frequency and Associated Factors

The number of patients undergoing metabolic bariatric surgery (MBS) has been increasing for many years [1, 2]. As obesity is a chronic and relapsing disease [3], patients may require secondary MBS due to unsatisfactory weight loss or complications from the initial procedure. Although MBS remains the most effective obesity treatment [4, 5], secondary surgery is becoming increasingly common, already accounting for 10–20% of cases in many countries [2]. While primary MBS is becoming more balanced in terms of finding the optimal treatment strategy, secondary procedures often yield less favorable results, and a significant variability in surgical approaches exists [6]. This highlights the need for focused research on secondary MBS.

Registry data, such as that from the Dutch Audit for Treatment of Obesity (DATO), can provide valuable insights into outcomes of secondary MBS. However, meaningful interpretation requires linkage with data from the primary procedure to put the outcomes of the secondary surgery in context, as both surgeries form part of the patient’s treatment. Such linkage is straightforward when both operations are performed in the same hospital, but becomes problematic when they occur at different institutions. Privacy regulations, imposed by the European General Data Protection Regulation (GDPR) [7], prohibit cross-institutional linkage, restricting analyses in DATO and other quality registries to patients treated within a single hospital. To assess the validity and representativeness of such analyses, it is essential to determine how often patients transfer to another hospital for secondary MBS and whether these patients differ systematically from those who remain at the same institution.

Addressing this limitation requires novel strategies for secure data linkage across institutions. A promising approach is secure multi-party computation (MPC), which enables collaborative analysis of sensitive information while preserving patient privacy. MPC relies on locally encrypting, fragmenting, and distributing the data across multiple independent servers hosted by a trusted third party (see Fig. 1), allowing analyses on the collectively processed data without exposing individual-level information to researchers or participating hospitals [8, 9]. Thereby, MPC facilitates the secure coupling of primary and secondary MBS data across different hospitals.

Bild vergrößern

Therefore, the aim of this study was to evaluate the frequency of hospital transfers for secondary MBS in the Netherlands using MPC and identify whether patients who transfer hospitals differ in patient profiles or the outcomes of their primary surgery compared with patients who remain in the same hospital for secondary MBS. Understanding these factors is critical for interpreting results of future research on secondary MBS and for improving the management of secondary MBS care.

In total, 98,409 encrypted surgical records were searched, and 2382 (2.4%) patients who received both primary and secondary MBS in the specified period were identified (see Fig. 2). Among these patients, 275 (11.5%) transferred to another hospital for the secondary procedure, while 2107 (88.5%) received the secondary surgery in the same hospital. Patients who transferred to another hospital had similar BMIs at primary surgery (43.9 vs 43.9, difference = 0.1, 95% confidence interval (CI) −0.8 to 0.9, p = 0.89), were younger (37.9 vs. 42.5, difference = 4.6, 95% CI 3.1–6.1, p < 0.001), and less often had hypertension or GERD (Table 1). They more frequently received a gastric bypass, while those staying in the same hospital more often had received sleeve gastrectomy (SG) as the primary procedure. Patients transferring hospitals more frequently received primary Roux-en-Y gastric bypass (RYGB) and less often one anastomosis gastric bypass (OAGB) (Table 1). The mean (SD) time between the two procedures was 4.1 (2.3) years for patients who transferred and 2.8 (2.1) years for those who stayed. At the time of secondary surgery, patients transferring to a different hospital were still younger, although the difference was smaller (42.0 vs. 45.4, difference = 3.4, 95% CI 1.9–4.8, p < 0.001), and had lower BMI (39.0 vs. 43.0, difference = 4.1, 95% CI 3.1–5.0, p < 0.001) (Table 2). The type of secondary surgery was mainly conversion surgery in both groups, but patients transferring hospitals more frequently underwent revision surgery (36.7% vs. 24.1%, difference = 12.6, 95% CI 0.6–24.6, p < 0.05). Among patients who transferred between hospitals, secondary surgery was more frequently performed for recurrent weight gain (RWG) (49.0% vs 23.0%, difference = 26.0, 95% CI 17.7–34.3, p < 0.001) and less often for GERD (4.8% vs 10.2%, difference = 5.4, 95% CI 1.6–9.2, p < 0.05). Patients who transferred hospitals less frequently experienced severe postoperative complications after the primary surgery (1.1% vs. 3.6%, difference = 2.5, 95% CI 1.0–4.0, p < 0.05).

Bild vergrößern

Using secure multi-party computation with real-world patient data, the proportion and characteristics of Dutch patients who transferred hospitals for secondary MBS and those who did not could be evaluated. Of all patients receiving both primary and secondary MBS between 2014 and 2022, one in nine transferred to another hospital for their secondary procedure. These patients were younger, had fewer obesity-related diseases, fewer severe postoperative complications, and more often received their secondary procedure for recurrent weight gain. The combination of a lower BMI at secondary surgery and a higher prevalence of recurrent weight gain as the indication may suggest that transferring patients are more easily dissatisfied with treatment outcomes, potentially leading them to seek second opinions; however, this requires further investigation.

While patient-initiated changes in healthcare providers are well-documented in various other medical fields [18‐20], prior MBS research has not studied this phenomenon. Reasons for hospital transfers in other fields may be to receive more specialized care, for convenience, e.g., moving to another city, or just general dissatisfaction with the care received [21‐23], suggesting that these factors may also influence MBS care decisions. Patients who transferred for secondary MBS were generally younger and had fewer obesity-related diseases, which may reflect age-related differences in healthcare behavior. A systematic review assessing second opinion patterns across 33 studies—albeit outside the MBS field—found that middle-aged patients, women, and those with higher education levels were more likely to seek a new care provider [24]. This aligns with a recent US survey of 20,000 patients, which reported that younger individuals were nearly six times more likely than older adults to change providers, primarily due to expectations of better care based on either personal experience or recommendations from others [25]. Only a minority (8%) cited logistical convenience, such as a more practical geographical location, as the reason for transferal. Those findings support the hypothesis that younger patients might be more critical or discerning regarding treatment outcomes. This is consistent with the current results, where transferring patients, despite having lower BMI at the time of secondary surgery, more often received secondary surgery for unsatisfactory weight loss. To assess the generalizability of these findings, similar analyses can be conducted in other countries to examine potential differences in patient and treatment characteristics associated with hospital transfer for secondary MBS.

Another insight from the results is that patients who stayed in their hospital more often had hypertension and GERD, potentially suggesting that patients might stay due to concurrent care provided by other specialists. The fact that the indication for secondary surgery was more often GERD in patients who stayed seems consistent with this explanation. Preoperative counselling on GERD after SG or OAGB, including the potential need for a conversional procedure to pursue resolution [26‐29], may reduce patient dissatisfaction and the likelihood of seeking treatment elsewhere. The current results support this theory, as patients who stayed were more likely to have received primary SG, and when receiving gastric bypass, more often underwent OAGB. This may be explained by the limited options for revision after RYGB, which are associated with either suboptimal additional weight loss (e.g., pouch resizing, limb length distalization, or ring augmentation) [30‐33] or high metabolic and nutritional complication rates (e.g., intestinal limb length alteration) [6, 34]. Meanwhile, SG can be converted relatively easily to RYGB, OAGB, or single anastomosis duodenal-ileal bypass with sleeve (SADI-S). Therefore, surgeons may be reluctant to offer secondary surgery for suboptimal weight loss after RYGB, leading to patients seeking a second opinion elsewhere [25]. Additionally, maintaining patient satisfaction requires preoperative counselling that addresses not only potential complications but also realistic expectations for weight loss. Because %TWL follows a normal distribution [35, 36], a proportion of patients will likely experience suboptimal outcomes. Patients should be made aware that they may fall at the lower end of this spectrum to prepare them for potentially unsatisfactory outcomes, rather than such results leading to dissatisfaction with their care.

Many countries have laws that protect personal health information to prevent misuse and encourage patients to be fully transparent during their consultation with healthcare professionals, ultimately benefiting patient care [7, 37, 38]. However, this legislation also hinders patient-data exchange for research that aims to improve patient care [39], and there is a need for pragmatic solutions. To our knowledge, MPC has been used with medical data from different data owners twice before [40, 41]. However, these studies were primarily designed as proof-of-principle experiments aimed at the technical aspects of MPC rather than applying the methodology to gain new insights. Additionally, only 192 and 48 patients were shared in those studies, resulting in relatively low computational demands. In contrast, nearly 100,000 surgical records were analyzed in the current study, thereby providing a robust test of the method’s capabilities. Given legal restrictions on sharing patient data between caregivers, MPC offers a solution for data analysis from multiple sources without requiring data exchange. Several well-established initiatives, such as EHDEN [42] and various Coordinated Registry Networks [43], are currently operational to facilitate data pooling across different institutions and countries. However, these projects do not aim to link shattered data from one patient but rather enable analyses on merged datasets to obtain new insights from larger data volumes. The current study serves as an example for other countries and study groups where researchers require data linkage but face constraints due to legislation or the reluctance of collaborators to exchange their sensitive data.

The results of this study benefit MBS providers by revealing that one in nine patients changes institutions for secondary surgery and that these patients tend to be younger with fewer obesity-related diseases. This insight enhances surgeons’ understanding of patient behavior and suggests that their secondary MBS rates, as computed using conventional DATO data, are likely slightly underestimated. The findings also highlight behavioral variability among patients, with younger individuals more commonly represented among patients transferring hospitals. This will only affect hospital performance comparisons if hospital transfers occur more frequently in some hospitals than in others, which seems unlikely. Surgeons may consider modifying their approach to secondary MBS to promote care continuity, given the potential for patients to pursue their desired treatment elsewhere. The study further demonstrates the utility of MPC, and potential collaborations could extend beyond hospitals to include insurers, governmental institutions, or other organizations with complementary data. For instance, obesity management medications (OMMs) are often prescribed by healthcare specialists not affiliated with the MBS institute, and MPC could facilitate analysis of the frequency and impact of (neo-)adjuvant OMM therapy. Beyond MBS, MPC can also be applied in other contexts, for instance, in cancer treatment, where patients often receive neoadjuvant therapy and surgery at different facilities.

The current study has several limitations. Reliance on registry data limited the availability of information on patient motivations, satisfaction, and decision-making, preventing evaluation of the hypothesis that transferring patients may be less satisfied with their outcomes or care. Additionally, outcomes after secondary surgery were not assessed, which would be relevant for determining whether hospital transfer is associated with better patient outcomes. However, this was beyond the study’s scope, as the primary aim was to compare patient and procedure characteristics to assess whether these differed systematically, but would be highly relevant for future research. Furthermore, no gold standard coupling key was available (e.g., social security number), leaving uncertain how many patients could have been linked, and how many patients were missed by coupling on the alternative patient identifier. It seems unlikely that, based on family name, biological sex, and date of birth, patients with certain characteristics would have a higher chance of being captured by the API than others. Therefore, it is not expected that this would have biased the results. Patients who transferred to a hospital abroad will also not be included in the current analysis, but this number is expected to be minimal. The current study also poses several challenges in data interpretation. Although the differences in characteristics between both groups may underlie the likelihood of hospital transfer, the assumption that younger, healthier patients are more likely to be unsatisfied with their treatment remains a hypothesis, and the actual reason for the transfer remains unknown. Additionally, the interval between the surgeries for patients who transferred hospitals was approximately 1 year longer, so the chance of RWG is larger. However, as the BMI of patients transferring hospitals was lower than that of patients who stayed, it does not change the assumption that patients who transferred were more critical of their outcomes. A limitation regarding the MPC software is that it mainly supports descriptive statistics but not regression analysis, which would allow controlling for other variables. In the context of the current study, this means that part of the observed differences may be due to another factor, e.g., the observed lower percentage of hypertension among patients who transferred to another hospital for secondary MBS may be due to these patients also being younger. Still, this does not essentially change the fact that patients who transfer to another hospital for secondary MBS are a selection of patients, reflected by several factors that might be interrelated.

This was the first time that MPC was applied to real-world data, analyzing approximately 100,000 medical records. The current study has helped in understanding whether secondary surgery outcomes based on patients remaining in their initial hospital, as currently done in DATO [44], could be generalizable to the broader MBS population. When future research evaluates the combined impact of primary and secondary MBS, it is critical to understand the potential bias introduced by missing data on patients who transferred hospitals. Since factors associated with weight loss, such as sex, BMI, and diabetes [4, 44], showed no baseline differences between patients who transferred and those who stayed, it suggests that weight loss results are likely representative. Still, future research should more broadly evaluate outcome differences between the two groups, such as complications and long-term outcomes after the secondary surgery. Additionally, it will be of interest to explore whether other factors are associated with hospital transfer, such as socioeconomic status or other characteristics that could be retrieved from third parties.

Approximately one in nine patients transferred to a different hospital for secondary MBS. These patients were younger, had lower rates of hypertension and GERD, and were more likely to have undergone gastric bypass, particularly RYGB, as their primary procedure. Despite presenting with lower BMI at the time of secondary MBS, they more frequently sought surgery for RWG. This combination suggests that younger patients may be more susceptible to dissatisfaction with their outcomes and more inclined to seek additional treatment elsewhere. The use of MPC facilitated these new insights, highlighting its significant potential for future research collaborations.