This study was approved by an independent IRB at the Faculty of Medicine, Chiang Mai University in Thailand. All the authors are certified in ICH-GCP training.
Translation of the CAM followed translation and cultural adaptation. First, NW translated the CAM into Thai, after which, a bilingual translator who had never been exposed to the original CAM did a backward translation. Then, both translators reached a consensus after comparing and adapting the test so that its meaning was fully congruent with Thai culture.
CAM training and inter-rater reliability
Assessing delirium using CAM requires training, and it should only be used with formal cognitive tests . To ensure rater reliability with the Thai version of CAM, monthly training sessions were set up for the participating family physicians. The training sessions included a summary of delirium and a detailed explanation of the Thai version of CAM, including various cognitive tests (such as MMSE Thai 2002 and a digit span test). Inter-rater reliability was ensured by asking the participating physicians to watch videos of ten delirious patients randomly selected from various wards. Three psychiatrist and family physician pairings were chosen at the beginning of the study, remaining in this pairing until the end, including during the training process. During the training, the authors focused only on the assessment of CAM specific items and on scoring instructions, rather than the expertise of the doctors involved. Thus, the training was based on how to use CAM and how to score each CAM item according to each patient's clinical manifestation. There was no clinical training regarding delirium given as part of the exercise. Disagreements regarding delirium diagnoses occurred between the pairs, especially at the beginning of the training (but not when diagnosing non-delirious patients). Most of these disagreements could be attributed to the family physicians' inability to detect poor attention (CAM item 2A) and disorganized thoughts (CAM item 3). The training, which included providing feedback, was repeated until, in the last ten random cases (five delirious; five non-delirious), each pair reached a 100% agreement on the diagnosis.
Participants and procedure
This study was a prospective validation study in 66 patients aged over 60 years of age newly admitted to a 2000-bed university-affiliated public hospital in Chiang Mai, Thailand. Over a five month period in 2009, a research nurse was responsible for screening and enrolling patients during the first 24 hours of their admission at the hospital. The nurse was given a list of each day's newly enrolled patients (with no diagnotic information included) and then randomly selected a small number of names each day from the list, for selection. After investigation, if the patients chosen turned out to have a previously diagnosed delirium condition, they were then excluded from taking part in the study. We also excluded patients admitted to the ICU with GCS < 8, or with significant hearing or visual impairments that would interfere with the testing process. For all patients chosen to take part, written informed consent was first received from the patient or their closest relative before proceeding.
The non-delirium experts used in this study were staff family physicians from the Department of Family Medicine at the Faculty of Medicine, Chiang Mai University, who spend most of their time carrying out general practitioner's work. On a day to day basis, they see patients within the setting of the primary care unit (PCU) at the hospital, under the jurisdiction of the Faculty of Medicine, as in Thailand, these PCUs are where residency training in Family Medicine takes place. On occasion, these physicians carry out outreach work within the community as part of an outreach community team. We invited these physicians to participate, as they represent a broad group of non-delirium experts, though they do play a role teaching residents how to detect delirium in the community.
MMSE Thai 2002  was performed by a research nurse to record the participants' intitial cognitive functioning, and subsequently each patient was assessed twice at their bedside within 72 hours of admission by two assessors: a psychiatrist and a family physician. In order to avoid bias due to fluctuations in altered consciousness, both examiners randomly assessed each patient within 30 minutes of each other. Furthermore, the family physicians and psychiatrists used in this research had had no prior experience taking care of the study patients on a day to day basis, and since the participants were randomly selected by a research assistant, the doctors were given no prior notice or information regarding the patients' conditions, whether they were delirious or otherwise. All the psychiatrists used in this study had at least ten years of experience assessing psychiatric patients with DSMs, and the diagnosis of delirium by psychiatrists can be regarded as the gold standard. The family physicians used the Thai version of the CAM algorithm to evaluate the presence or not of delirium, and at the end of each interview, the examiners independently recorded their diagnoses. All the interviews were videotaped for both training and post-analysis purposes.
Cohen's Kappa was used to calculate the extent of agreement between the family physicians and the psychiatrists, in terms of their diagnoses. The sensitivity and specificity of the Thai version of the CAM algorithm, as used by the family physicians, were calculated and compared to the diagnoses given by the psychiatrists using DSM-IV TR. Demographic variables and descriptive statistics (i.e. percentages) were reported, and a non-parametric statistical analysis was applied to data without a normal distribution, and in addition, the positive predictive value (PPV) and negative predictive value (NPV) were both determined. The Statistical Package for the Social Sciences (SPSS 17.0 for Windows) was used.