After decades of decrease, maternal mortality rates have shown a slight increase in Europe. Immigrants, especially Africans, have shown to be at higher risk than native women. This could not be explained solely by well-known obstetric and socio-economic risk factors. The aim of this thesis was to study incidence, classification and quality of care of maternal deaths in Sweden, with focus on the foreign-born population. The study population was identified through linkage of the Cause of Death Register, Medical Birth Register, and National Patient Register, and medical records obtained from hospitals. Data from registers, death certificates, and medical records were reviewed. Suboptimal care was studied by structured implicit review of medical records. Differences between foreign- and Swedish-born women were analysed by relative risks, Chi2- and Fisher’s exact test. Underreporting of maternal mortality was shown to be substantial: as compared to the official statistics, 64% more maternal deaths were identified. Women born in low-income countries were identified as being at highest risk of dying during reproductive age in Sweden. The relative risk of dying from diseases related to pregnancy was 6.6 (95% confidence interval 2.6–16.5) for women born in low-income countries, as compared to Swedish-born women. Major and minor suboptimal factors related to care-seeking, accessibility, and quality of care were found to be associated with a majority of maternal deaths and significantly more often to foreign-born women. Suboptimal factors identified included non-compliance, communication barriers, and inadequate care. The rate of suicides during pregnancy or within one year after delivery did not change during the last three decades, and was higher for foreign-born women. A majority of women who committed suicide had been under psychiatric care, but such documentation at antenatal care was inconsistent, and planning for follow-up postpartum was generally lacking. The conclusion of this thesis is that foreign-born women are a high-risk group for maternal death and morbidity that calls for clinical awareness with respect to their somatic and psychiatric history, care-seeking behaviour, and communication barriers. Cross-disciplinary care is necessary, both in obstetric emergencies and in cases of maternal psychiatric illness, to avert maternal death and suicide.

L’objectif de cette thèse était d’étudier la distribution, les déterminants et les conséquences des délais diagnostiques des cancers de l’enfant, par une revue systématique de la littérature, une analyse des plaintes déposées en France et au Canada et par des études ad hoc en population sur les tumeurs ayant des délais particulièrement longs : le médulloblastome et le sarcome d’Ewing.La revue systématique n’a pas retrouvé de diminution significative de la longueur des délais au cours du temps. Les délais longs étaient associés à un âge élevé, au type histologique et à la localisation de la tumeur. Les relations entre délai et gravité de la maladie étaient variables. Seul un tiers des conclusions des expertises judiciaires (n = 56) étaient concordantes avec les données de la littérature.Le délai diagnostique médian des enfants atteints de médulloblastome en Ile-de-France (n = 166) était de 65 jours. Les délais longs étaient associés paradoxalement à des métastases moins fréquentes et à une histologie favorable, mais pas à la survie ni aux séquelles.Le délai diagnostique médian des enfants atteints de sarcome d’Ewing en France (n = 436) était de 70 jours. Les délais longs, liés à un âge élevé et à la localisation de la tumeur, n’étaient pas associés au volume tumoral, à la présence de métastases, à l’opérabilité ni à la survie.Pour certaines tumeurs, une association entre délais diagnostiques et gravité est établie (rétinoblastome) ou hautement probable. Pour d’autres, l’absence d’association démontrée permettrait de dédramatiser la perception de leurs conséquences, sans dispenser d’essayer de réduire ces délais, notamment pour en atténuer les conséquences psychologiques. / The aim of this thesis was to study the distribution, determinants and consequences of time to diagnosis of cancer in children, through a systematic review of the literature and an analysis of lawsuits in France and Canada, and two population-based studies of tumors with particularly long diagnosis delays: medulloblastoma and Ewing sarcoma.The systematic review did not identify any significant decreases in time to diagnosis during the studies. Long times to diagnosis were associated with older age, histological type and location of the tumor. Associations between time to diagnosis and severity of the disease varied. Only a third of the court-appointed experts (n = 56) provided testimony concordant with the available medical literature.The median time to diagnosis of children with medulloblastoma in the area of Paris (n = 166) was 65 days. Diagnosis delays were paradoxically associated with less frequent metastasis and favorable histology, but not with survival, or sequelae.The median time to diagnosis of children with Ewing sarcoma in France (n = 436) was 70 days. Diagnosis delays, related with older age and tumor location, were not associated with tumor size, presence of metastasis, surgical outcome, or survival.For some tumors, an association between time to diagnosis and severity of the disease is well established (e.g. retinoblastoma), or highly probable. For others, the lack of demonstrated associations could tone down the perception of the supposed consequences of diagnosis delays - but does not exempt from trying to reduce them, in order to alleviate their psychological consequences.

Les objectifs de cette thèse étaient de produire des connaissances nouvelles sur les méthodologies d’évaluation de l’optimalité des soins avec l’exemple des délais diagnostiques et des infections bactériennes sévères de l’enfant (IBS). Nous avons mis en évidence, dans deux revues systématiques de la littérature, que les points méthodologiques potentiellement associés à des risques de biais et d’obstacles à la transportabilité des résultats étaient rarement rapportés dans les études primaires sur les délais diagnostiques ou rarement évalués par les auteurs de méta-analyses. Nous avons donc construit et validé internationalement une reporting guideline pour aider les scientifiques à prendre en compte ces points méthodologiques critiques. Nous avons montré par une enquête confidentielle avec comité d’experts en population que : (i) les prises en charge étaient suboptimales pour 76% des enfants décédés d’IBS, (ii) un retard au recours médical, une sous-évaluation de la gravité ou un retard à l’antibiothérapie étaient retrouvés dans la prise en charge de respectivement 20%, 20% et 24% des enfants atteints d’IBS, (iii) les soins suboptimaux étaient indépendamment et fortement associés au risque de décès et (iv) les soins suboptimaux étaient plus fréquents chez les enfants de moins d’un an ou lorsque qu’ils n’étaient pas administrés par un médecin spécifiquement formé. La minimisation des biais dans la sélection des participants et la mesure de l’optimalité et la prise en compte de facteurs de confusion comme la sévérité intrinsèque de la maladie sont des éléments clefs de l’évaluation de l’optimalité des soins afin de produire des messages cliniques correctifs valides. / The aim of this thesis was to product new knowledge about the methodology on how to assess the optimality of care with the examples of time to diagnosis and serious bacterial infection (SBI). In two systematic reviews, we found that the key methodological points potentially related to risks of bias or threats to transportability were rarely reported in the primary studies and rarely evaluated by authors of systematic reviews. Then, we developed and internationally validated a reporting guideline to help scientists to better take into consideration these critical methodological points. In a population-based confidential inquiry, we found that: (i) care was suboptimal in 76% of the initial management of children who died from SBI, (ii) delayed first medical contact, undervaluation of severity or delayed antibiotic administration were detected in the management of 20%, 20% and 24% of children admitted to intensive care for a SBI, respectively, (iii) the total number of suboptimal cares delivered during the management was independently associated with death, and (iv) suboptimal cares were more frequent in children younger than one year old and if the care was delivered by a non specialist physician. Minimizing the risks of bias both in the selection process of the study population and in the assessment of the optimality of care, and taking into account confounding factors such as the intrinsic severity of the disease are keys elements to ensure a reliable evaluation of optimality of care in order to produce effective corrective clinical messages.