Évaluation des soins et surveillance des maladies cardiovasculaires: pouvons-nous faire confiance aux données médico-administratives hospitalières? / Cardiovascular disease assessment and surveillance: can we trust hospital medical and administrative data?

INTRODUCTION: L'Institut national de santé publique du Québec (INSPQ) a reçu le mandat du ministère de la Santé et des Services sociaux (MSSS) de développer la surveillance des maladies cardiovasculaires au Québec, et l'Institut national d'excellence en santé et en services sociaux (INESSS), le mandat d'évaluer des technologies et des modes d'intervention en cardiologie. L'évaluation des soins et la surveillance des maladies constituent des aspects essentiels d'un système de santé moderne et sont d'une importance capitale en regard des maladies cardiovasculaires, ces dernières étant très répandues et associées à des coûts considérables. Les deux activités reposent souvent sur l'utilisation de fichiers médico-administratifs. Or, les données médico-administratives offrent non seulement une couverture complète de la population, mais sont également mises à jour continuellement. La base de données hospitalières du Québec, MED-ÉCHO (Maintenance et Exploitation des Données pour l'Étude de la Clientèle HOspitalière), fournit un profil démographique et clinique du patient à un coût minime. D'une part, cette base a l'avantage théorique d'intégrer des données propres aux antécédents médico-hospitaliers (informations obtenues à partir d'hospitalisations antérieures) susceptibles d'améliorer l'ajustement du risque en déterminant les comorbidités absentes lors de l'admission de référence ainsi que l'ajout d'informations sur la durée des comorbidités chroniques. D'autre part, la qualité des données médico-administratives hospitalières repose sur des données sommaires documentées par les médecins et codées par les archivistes médicaux, ce qui peut ainsi entraîner d'importantes limitations à l'égard des erreurs de classification des comorbidités, de la cohérence clinique et de la distinction entre des comorbidités préexistantes et des complications survenant lors de l'hospitalisation de référence. De toute évidence, l'évaluation des soins et la surveillance des maladies seraient considérablement simplifiées si les données médico-administratives existantes étaient suffisamment fiables et que les analyses tirées de ces données étaient crédibles aux yeux des cliniciens. Conséquemment, la présente étude a pour but de répondre à trois questions: 1. Les données de MED-ÉCHO sont-elles assez fiables, au sein d'une cohorte de patients hospitalisés atteints de maladies cardiovasculaires (aux prises avec un infarctus aigu du myocarde (IAM) ou ayant subi une des interventions de revascularisation), pour déterminer les cas d'IAM et les comorbidités qui y sont le plus souvent associées, en comparaison avec les informations indiquées dans les dossiers médicaux hospitaliers? 2. Les modèles de prédiction de la mortalité, fondés sur les comorbidités retrouvées dans MED-ÉCHO, sont-ils comparables à ceux qui s'appuient sur les comorbidités repérées au moyen de la revue des dossiers médicaux? 3. Quelle est la valeur ajoutée de l'intégration des données propres aux antécédents médicaux de MED-ÉCHO dans le modèle qui utilise les données médico-administratives de l'admission de référence? MÉTHODES: Treize centres hospitaliers ont été sélectionnés dans la province dans le but de créer un échantillon représentatif de: 1. Différents types de centres hospitaliers québécois (centres primaires : absence de coronarographie, d'angioplastie ou de chirurgie cardiaque ; centres secondaires : coronarographie et angioplastie, mais absence de chirurgie cardiaque ; centres tertiaires : coronarographie et angioplastie ainsi que chirurgie cardiaque); 2. Divers volumes hospitaliers; et 3. Différentes régions sociosanitaires. Les centres hospitaliers primaires et secondaires sont distribués dans sept régions sociosanitaires, alors que les cinq centres hospitaliers tertiaires sont répartis dans trois régions. Sélection des patients; Sélection des comorbidités; Révision des dossiers médicaux; Variabilité inter-archivistes de la révision des dossiers médicaux hospitaliers; Données médico-administratives (MED-ÉCHO); Analyses statistiques; CONCLUSION: Les principales conclusions de cette étude, menée à l'aide d'une cohorte de patients cardiaques hospitalisés dans un échantillon représentatif de centres hospitaliers québécois, se résument ainsi: 1. La présence de comorbidités dans MED-ÉCHO se retrouve presque intégralement dans les dossiers médicaux hospitaliers; 2. Un modèle prédictif de la mortalité à un an, basé sur les données de MED-ÉCHO, se compare très favorablement avec un même modèle prédictif fondé sur la révision des dossiers médicaux hospitaliers; 3. La possibilité d'intégrer l'information relative aux hospitalisations précédentes en plus de l'information obtenue quant à l'admission de référence dans les données de MED-ÉCHO améliore la détection des comorbidités. Ce dernier point est essentiel, car chez les patients ayant eu une courte hospitalisation, voire même un séjour hospitalier de type « urgent ¼ et bref (ce qui est souvent le cas avec l'ICP), la possibilité d'obtenir les comorbidités propres aux hospitalisations précédentes peut présenter un avantage de taille pour ce qui est des données médico-administratives. Dans notre étude, l'utilité possible des données propres aux antécédents médico-hospitaliers a été bien illustrée par le fait que la prévalence d'un IAM ancien, selon la revue des dossiers médicaux hospitaliers, était de 21 %, tandis qu'elle était de 25 % avec la présence d'un tel code dans MED-ÉCHO pour au moins une hospitalisation dans l'année précédant l'admission de référence, comparativement à 12 % si seule l'admission de référence dans MED-ÉCHO était prise en considération. Ainsi, les données administratives propres aux antécédents médico-hospitaliers peuvent être une source d'information plus fiable que les dossiers médicaux hospitaliers pour déterminer les antécédents médicaux d'un patient, surtout si le patient a été hospitalisé dans différents établissements. Pour terminer, mentionnons que nos résultats suggèrent que la banque de données médico-administrative hospitalière MED-ÉCHO est, en général, bien codée et se compare favorablement aux résultats obtenus à la révision des dossiers médicaux hospitaliers pour ce qui est de la prédiction de la mortalité. La capacité prédictive de la banque de données MED-ÉCHO semble également améliorée grâce à l'ajout de données propres aux antécédents médico-hospitaliers. Elle peut donc être utilisée avec confiance pour informer les cliniciens et le public à propos des résultats cliniques et pour déterminer les problèmes de soins de santé qui justifient des recherches plus approfondies.

Summary: computed tomography and Its carcinogenic effects in children and youth in Québec

Developments in diagnostic radiology have made it possible not only to perform increasingly precise scans and but also to reduce their associated radiation doses. Yet the growing use of these scans in medical practice and the advent of new technologies expose the population to an increasingly high collective dose. Computed tomography (CT) has been estimated to account for roughly 50% of diagnostic medical radiation in the general public, with another 25% due to nuclear medicine, and the remaining 25% due to all the other radiological tests (conventional radiography, fluoroscopy, PET, etc.) [Fazel et al., 2009; Mettler et al., 2008a]. This report was prepared at the request of the Direction québécoise du cancer (DQC), which was especially interested in several issues: the cancer-related risks of diagnostic CT in children and young adults; available imaging options; and the means taken to reduce CT-related risks. The DQC asked us to focus on children and youth, given their increased sensitivity to radiation, their longer life expectancy, and the possibility that absorbed doses may be higher in these patients. The aspects studied concerned the link between cancer and radiation from diagnostic medical procedures. Our examination of the link between radiation and cancer development, and of the quantification of absorbed doses (dosimetry), raised the following questions: What is the best estimate of the extent of CT use in children and youth in Québec? What information is there on (i) the number of scans per age group; (ii) the effective dose associated with these scans; (iii) the contributions of the different CT scans; and (iv) radiation doses delivered in CT compared with those from other radiation sources? What is the estimated carcinogenesis, specifically, the number of new cancer cases (incidence) and deaths (mortality) associated with CT in children and youth below the age of 20 years in Québec? This review summarizes the literature on the link between ionizing radiation and cancer. It presents relevant concepts and measurement units, along with epidemiological evidence on the causal link between radiation and cancer. To clearly highlight the areas of consensus or debate, we performed a literature search in the databases containing primary scientific articles and systematic reviews published from 2000 to 2010, with no restriction as to language of publication. Special attention was paid to studies addressing CT-related radiation in children and youth below the age of 20 years. An estimate of different exposures to natural radiation sources is provided based on information available worldwide, in Canada and in Québec. A similar estimate was produced for artificial radiation sources, primarily from medical procedures. We used data supplied by the RAMQ for 2009 and a dose-effect relationship model published in the literature to estimate the number of new cancer cases and cancer-related deaths potentially occurring during the lifetime of those exposed to CT radiation. These estimates were compared with others previously produced for Québec, the rest of Canada and the United States. It was found that less than 4% of the CT scans done in Québec in 2009 were performed in children and youth aged 0 to 19 years, from a total of approximately 888,000 scans. This percentage varied with the anatomical site examined. Data indicate that 35% of CT procedures in adults were abdominal (26% for abdomen and pelvis, 8% for chest and abdomen, 1% for abdomen alone), whereas the head was the most common area of investigation in children and youth below the age of 20 years. No matter how small, the risk associated with the thousands of scans performed each year suggests that a certain number of cancers are likely to occur. By applying a risk model adjusted for age distribution and for the health status of people undergoing CT scans, Berrington de González et al. [2009] derived an estimate of 14,500 deaths per year in the U.S. (equal to approximately 360 deaths in Québec). An analogous calculation was applied here to derive a more appropriate estimate for Québec in 2009, based on information on CT scan volume per anatomical site and per age-sex distribution. For estimating carcinogenesis in children, the number of scans and the effective doses associated with different pediatric CT scans were used under the basic assumption that pediatric doses are similar to adult doses. In 2009, 1.7 million children and youth below the age of 20 years accounted for 22% of the Québec population. The 32,668 CT scans in these children and youth accounted for 3.7% of those performed for all ages, and for 2.5% of the effective doses associated with these scans. The link between radiation and cancer is well established, but the exact quantification of its effects remains a topic of controversy, especially for doses less than 100 mSv, which are typical in medical practice. For our estimates, we used the BEIR model [NRC, 2006], which assumes a linear relationship between the doses received and the probability of developing cancer. The application of these doses to a patient population may overestimate the carcinogenic effect produced in people who may sometimes have limited life expectancies. Berrington de González et al. [2009] reduced their cancer risk estimate by 11%, accounting for people who die from cancer within five years of their scans, and by 9% by eliminating the scans performed in people who had already been diagnosed with cancer. By applying the 20% reduction proposed by Berrington de González, we obtained an estimate on the order of 300 new cancer cases (estimate produced with our model: 286 new cancer cases and 184 deaths), including twenty or so in children and youth below the age of 20 years (estimate produced with our model: 24 cancer cases and 11 deaths in children and youth below the age of 20 years). By comparison, we estimated that in 2009 there would have been 44,200 new cancer cases in Québec and 20,100 cancer deaths, accounting for 35% of the 57,200 deaths from all causes in Québec. Changing demographics will cause these figures to rise rapidly over the next two or three decades. The number of potential CT-induced cancers in Québec therefore accounts for only a small proportion of the total number of cancers. Estimates of the number of cancer cases and the number of cancer-related deaths are dependent upon various assumptions, and reasonable variations in these assumptions may easily generate estimates roughly twice as high or twice as low. This quantitative analysis considered only the negative effects of CT imaging, which may be quantified by estimation. However, it is much more difficult to measure the morbidity and mortality that these 32,668 CT scans may have prevented, owing to earlier or more accurate diagnoses of serious diseases. In the case of CT scans, the conventional medical attitude is to expect positive effects to significantly outweigh negative effects. This attitude should perhaps be tempered by recognizing the cancer risk associated with these scans, while emphasizing the importance of the diagnostic information they yield. The challenge of implementing best practices in CT is to make sure that this is the case for each clinical indication for which it is used. Two major strategies may help ensure the preponderance of positive effects: optimization of CT performance and justification of the use of CT scans. Internationally, the risk associated with diagnostic radiation has been recognized since the advent of radiology and nuclear medicine. Historically, risk control, through radiation protection and the judicious use of these technologies, has kept pace with developments in the available techniques. The advent of CT, which delivers much higher radiation doses than conventional radiography, has led to new initiatives for aligning practices with this new reality, especially for pediatric patients. Examples include the Image Gently movement launched in 2006, the FDA's collaborative initiatives with manufacturers early in 2010, and Health Canada's Security Code 35 introduced in 2009. The present analysis provides an overview of the measures found in the medical literature that are designed to limit radiation-related dangers, while preserving the many benefits that these technologies offer to patients. These measures include: -providing professional training and awareness about the issues; -producing information to help physicians make decisions about the use of CT scans; -developing dose-rate standards; -developing guidelines on the indications for the different imaging techniques; and -implementing quality-assurance measures in clinical and radiology facilities. In this dynamic environment, these measures must be developed and supported to allow patients to take full advantage of the promising benefits of these technologies and to reduce the risk of inducing cancers that is inevitably linked to ionizing radiation.(AU)

Screening mammography: a reassessment

INTRODUCTION: Eight trials examining the performance of screening mammography have been conducted in the USA, Sweden, the United Kingdom and Canada, beginning in 1963. A first report by the Conseil d'évaluation des technologies de la santé (CETS) published in 1990 concluded that screening mammography trials had shown reductions in mortality from breast cancer of 35%, with 45% in the subgroup of women aged 50 to 69. A second report in 1993 concluded that mammographic screening of women under 50 had not been shown to reduce mortality. By the year 1998, when Québec introduced the Programme québécois de dépistage du cancer du sein (PQDCS), all Canadian provinces and many other countries had organized screening programs in place. A recent Cochrane Collaboration Group review, challenging the belief that mammography screening is an effective tool for reducing breast cancer deaths, has raised concerns about the validity of the published randomized trials. This update addresses three questions: (1) What is the strength of the scientific evidence on which screening mammography programs are based? (2) What is the evidence in support of screening for women aged 40 to 49 years? (3) What are the implications of research studies for maximizing the effectiveness of modern programs such as the Programme québécois de dépistage du cancer du sein (PQDCS)? METHODOLOGIC ANALYSIS: An evaluation of efficacy trials essentially aims to determine whether the conditions under which the trials were performed and the results that were obtained can guide strategies. In practice, the reference strategy (no screening) may include some uncontrollable screening activities, which will weaken the contrast with the screening intervention. A valid study must be a fair comparison between screening and no screening. Thus screening and control cohorts should have the same baseline risk of breast cancer mortality, should be treated equally in all regards except concerning the screening or control intervention, and should have the information on their outcome measured in a way that is independent of their assignment to the screening or control group. Validity can be compromised by bias of known direction and by bias of unknown direction. In this evaluation, to further develop the notion of bias of known direction, we use the concept of strength of contrast. It corresponds to the degree to which a trial succeeds in bringing out the divergence between the two strategies compared and in measuring the effects that this divergence produces. Five elements are evaluated in this report which help assess the strength of contrast: -the technical contrast, or the nature of the difference between screening and control interventions; -the era in which these techniques are applied; -the quality of the intervention, including quality control measures; -rates of participation and contamination measured among screening and control cohorts; and -the timing of the measurement of the effects of screening on mortality (or timing dilution). DISCUSSION AND CONCLUSIONS: Question 1: What is the strength of the scientific evidence on which screening mammography programs are based? There are serious concerns regarding the validity of most of the trials supporting mammography screening, based on methodological weaknesses in the screening trials. Studies are highly heterogeneous with regard to the strength of the contrast that they studied, with numerous weaknesses identified in all the major studies, meaning that the potential of screening mammography has perhaps not been thoroughly explored. Using the best available data, one can conclude that there is fair evidence of moderate reduction of breast cancer mortality, of the order of 9 to 15%; data restricted to women over the age of 50 show greater reductions, of the order of 24 to 29%. Furthermore, our analysis has demonstrated that modern mammography, carried out under quality conditions that maximize its performance, has the potential to identify cancerous lesions earlier in their progression, and this may allow for some further reduction in mortality. Conclusion: Existing scientific trials, despite their flaws, support mammography screening programs. In addition, there are good reasons to believe that modern, well-conducted screening programs may achieve earlier detection and diagnosis of breast cancer and, perhaps, greater reductions in breast cancer mortality than what has been found in screening trials. Question 2: What is the evidence in support of screening mammography for women aged 40 to 49 years? There is much less data available to answer the question, since most study experience is in women over 50, even though some women in some of the studies started screening several years earlier than their fiftieth birthday. The best data available show no significant reduction in breast cancer mortality in women screened before the age of 50. In the absence of any convincing data that mammography is efficacious in this age group, harmful effects may outweigh any positive effects. Conclusion: Trial data published to date do not provide scientific justification to recommend screening for women younger than 50. However, this conclusion does not exclude the possibility that screening of individual women, based on a personalized risk assessment, could be of benefit. These conclusions should be reviewed when results from the UK Trial become available. Question 3: What are the implications of research studies for maximizing the effectiveness of modern programs such as the Programme québécois de dépistage du cancer du sein (PQDCS)? Although the PQDCS already includes rigorous control of the quality of films produced, certain aspects of the structure and process of trials examined under the rubric of strength of contrast can be transposed as additional quality norms. Notable among these are double reading of films and an annual reading volume sufficient to allow each radiologist to acquire and maintain the necessary expertise to detect breast cancer in its early stages. These aspects should also allow for a reduction in false positive rates and subsequent unnecessary diagnostic procedures. Moreover, high participation rates at each screening round will contribute to achieving and perhaps exceeding the mortality reductions obtained by screening trials. Conclusion: Modern screening programs such as the PQDCS may produce outcomes comparable or even superior to those observed in screening trials if they achieve a standard of quality equal to or better than the standard achieved by trials. Measures that should reduce false positive rates and assure high-quality screening include making sure that high-quality mammographic films are being produced, that readers have the necessary expertise to detect early cancer and avoid false positives, and double reading of a proportion of films. While participation rates should be as high as possible, efforts to increase participation should not overstate the benefits of mammography nor understate the risks and uncertainties which remain.