Econstudentlog

Data on Danish diabetics (Dansk Diabetes Database – National årsrapport 2013-2014)

[Warning: Long post].

I’ve blogged data related to the data covered in this post before here on the blog, but when I did that I only provided coverage in Danish. Part of my motivation for providing some coverage in English here (which is a slightly awkward and time consuming thing to do as all the source material is in Danish) is that this is the sort of data you probably won’t ever get to know about if you don’t understand Danish, and it seems like some of it might be worth knowing about also for people who do not live in Denmark. Another reason for posting stuff in English is of course that I dislike writing a blog post which I know beforehand that some of my regular readers will not understand. I should perhaps note that some of the data is at least peripherally related to my academic work at the moment.

The report which I’m covering in this post (here’s a link to it) deals primarily with various metrics collected in order to evaluate whether treatment goals which have been set centrally are being met by the Danish regions, one of the primary political responsibilities of which is to deal with health care service delivery. To take an example from the report, a goal has been set that at least 95 % of patients with known diabetes in the Danish regions should have their Hba1c (an important variable in the treatment context) measured at least once per year. The report of course doesn’t just contain a list of goals etc. – it also presents a lot of data which has been collected throughout the country in order to figure out to which extent the various goals have been met at the local levels. Hba1c is just an example; there are also goals set in relation to the variables hypertension, regular eye screenings, regular kidney function tests, regular foot examinations, and regular tests for hyperlipidemia, among others.

Testing is just one aspect of what’s being measured; other goals relate to treatment delivery. There’s for example a goal that the proportion of (known) type 2 diabetics with an Hba1c above 7.0% who are not receiving anti-diabetic treatment should be at most 5% within regions. A thought that occurred to me while reading the report was that it seemed to me that some interesting incentive problems might pop up here if these numbers were more important than I assume they are in the current decision-making context, because adding this specific variable without also adding a goal for ‘finding diabetics who do not know they are sick’ – and no such goal is included in the report, as far as I’ve been able to ascertain – might lead to problems; in theory a region that would do well in terms of identifying undiagnosed type 2 patients, of which there are many, might get punished for this if their higher patient population in treatment as a result of better identification might lead to binding capacity constraints at various treatment levels; capacity constraints which would not affect regions which are worse at identifying (non-)patients at risk because of the existence of a tradeoff between resources devoted to search/identification and resources devoted to treatment. Without a goal for identifying undiagnosed type 2 diabetics, it seems to me that to the extent that there’s a tradeoff between devoting resources to identifying new cases and devoting resources to the treatment of known cases, the current structure of evaluation, to the extent that it informs decision-making at the regional level, favours treatment over identification – which might or might not be problematic from a cost-benefit point of view. I find it somewhat puzzling that no goals relate to case-finding/diagnostics because a lot of the goals only really make sense if the people who are sick actually get diagnosed so that they can receive treatment in the first place; that, say, 95% of diabetics with a diagnosis receives treatment option X is much less impressive if, say, a third of all people with the disease do not have a diagnosis. Considering the relatively low amount of variation in some of the metrics included you’d expect a variable of this sort to be included here, at least I did.

The report has an appendix with some interesting information about the sex ratios, age distributions, how long people have had diabetes, whether they smoke, what their BMIs and blood pressures are like, how well they’re regulated (in terms of Hba1c), what they’re treated with (insulin, antihypertensive drugs, etc.), their cholesterol levels and triglyceride levels, etc. I’ll talk about these numbers towards the end of the post – if you want to get straight to this coverage and don’t care about the ‘main coverage’, you can just scroll down until you reach the ‘…’ point below.

The report has 182 pages with a lot of data, so I’m not going to talk about all of it. It is based on very large data sets which include more than 37.000 Danish diabetes patients from specialized diabetes units (diabetesambulatorier) (these are usually located in hospitals and provide ambulatory care only) as well as 34.000 diabetics treated by their local GPs – the aim is to eventually include all Danish diabetics in the database, and more are added each year, but even as it is a very big proportion of all patients are ‘accounted for’ in the data. Other sources also provide additional details, for example there’s a database on children and young diabetics collected separately. Most of the diabetics which are not included here are patients treated by their local GPs, and there’s still a substantial amount of uncertainty related to this group; approximately 90% of all patients connected to the diabetes units are assumed at this point to be included in the database, but the report also notes that approximately 80 % of diabetics are assumed to be treated in general practice. Coverage of this patient population is currently improving rapidly and it seems that most diabetics in Denmark will likely be included in the database within the next few years. They speculate in the report that the inclusion of more patients treated in general practice may be part of the explanation why goal achievement seems to have decreased slightly over time; this seems to me like a likely explanation considering the data they present as the diabetes units in general are better at achieving the goals set than are the GPs. The data is up to date – as some of you might have inferred from the presumably partly unintelligible words in the parenthesis in the title, the report deals with data from the time period 2013-2014. I decided early on not to copy tables into this post directly as it’s highly annoying to have to translate terms in such tables; instead I’ve tried to give you the highlights. I may or may not have succeeded in doing that, but you should be aware, especially if you understand Danish, that the report has a lot of details, e.g. in terms of intraregional variation etc., which are excluded from this coverage. Although I far from cover all the data, I do cover most of the main topics dealt with in the publication in at least a little bit of detail.

The report concludes in the introduction that for most treatment indicators no clinically significant differences in the quality of the treatment provided to diabetics are apparent when you compare the different Danish regions – so if you’re looking at the big picture, if you’re a Danish diabetic it doesn’t matter all that much if you live in Jutland or in Copenhagen. However some significant intra-regional differences do exist. In the following I’ll talk in a bit more detail about some of data included in the report.

When looking at the Hba1c goal (95% should be tested at least once per year), they evaluate the groups treated in the diabetes units and the groups treated in general practice separately; so you have one metric for patients treated in diabetes units living in the north of Jutland (North Denmark Region) and you have another group of patients treated in general practice living in the north of Jutland – this breakdown of the data makes it possible to not only compare people across regions but also to investigate whether there are important differences between the care provided by diabetes units and the care provided by general practitioners. When dealing with patients receiving ambulatory care from the diabetes units all regions meet the goal, but in Copenhagen (Capital Region of Denmark, (-CRD)) only 94% of patients treated in general practice had their Hba1c measured within the last year – this was the only region which did not meet the goal for the patient population treated in general practice. I would have thought beforehand that all diabetes units would have 100% coverage here, but that’s actually only the case in the region in which I live (Central Denmark Region) – on the other hand in most other regions, aside from Copenhagen again, the number is 99%, which seems reasonable as I’m assuming a substantial proportion of the remainder is explained by patient noncompliance, which is difficult to avoid completely. I speculate that patient compliance differences between patient populations treated at diabetes units and patient populations treated by their GP might also be part of the explanation for the lower goal achievement of the general practice population; as far as I’m aware diabetes units can deny care in the case of non-compliance whereas GPs cannot, so you’d sort of expect the most ‘difficult’ patients to end up in general practice; this is speculation to some extent and I’m not sure it’s a big effect, but it’s worth keeping in mind when analyzing this data that not all differences you observe necessarily relate to service delivery inputs (whether or not a doctor reminds a patient it’s time to get his eyes checked, for example); the two main groups analyzed are likely to also be different due to patient population compositions. Differences in patient population composition may of course also drive some of the intraregional variation observed. They mention in their discussion of the results for the Hba1c variable that they’re planning on changing the standard here to one which relate to the distributional results of the Hba1c, not just whether the test was done, which seems like a good idea. As it is, the great majority of Danish diabetics have their Hba1c measured at least annually, which is good news because of the importance of this variable in the treatment context.

In the context of hypertension, there’s a goal that at least 95% of diabetics should have their blood pressure measured at least once per year. In the context of patients treated in the diabetes units, all regions achieve the goal and the national average for this patient population is 97% (once again the region in which I live is the only one that achieved 100 % coverage), but in the context of patients treated in general practice only one region (North Denmark Region) managed to get to 95% and the national average is 90%. In most regions, one in ten diabetics treated in general practice do not have their blood pressure measured once per year, and again Copenhagen (CRD) is doing worst with a coverage of only 87%. As mentioned in the general comments above some of the intraregional variation is actually quite substantial, and this may be a good example because not all hospitals are doing great on this variable. Sygehus Sønderjylland, Aabenraa (in southern Jutland), one of the diabetes units, had a coverage of only 67%, and the percentage of patients treated at Hillerød Hospital in Copenhagen (CRD), another diabetes unit, was likewise quite low, with 83% of patients having had their blood pressure measured within the last year. These hospitals are however the exceptions to the rule. Evaluating whether it has been tested if patients do or do not have hypertension is different from evaluating whether hypertension is actually treated after it has been discovered, and here the numbers are less impressive; for the type 1 patients treated in the diabetes units, roughly one third (31%) of patients with a blood pressure higher than 140/90 are not receiving treatment for hypertension (the goal was at most 20%). The picture was much better for type 2 patients (11% at the national level) and patients treated in general practice (13%). They note that the picture has not improved over the last years for the type 1 patients and that this is not in their opinion a satisfactory state of affairs. A note of caution is that the variable only includes patients who have had a blood pressure measured within the last year which was higher than 140/90 and that you can’t use this variable as an indication of how many patients with high blood pressure are not being treated; some patients who are in treatment for high blood pressure have blood pressures lower than 140/90 (achieving this would in many cases be the point of treatment…). Such an estimate will however be added to later versions of the report. In terms of the public health consequences of undertreatment, the two patient populations are of course far from equally important. As noted later in the coverage, the proportion of type 2 patients on antihypertensive agents is much higher than the proportion of type 1 diabetics receiving treatment like this, and despite this difference the blood pressure distributions of the two patient populations are reasonably similar (more on this below).

Screening for albuminuria: The goal here is that at least 95 % of adult diabetics are screened within a two-year period (There are slightly different goals for children and young adults, but I won’t go into those). In the context of patients treated in the diabetes units, the northern Jutland Region and Copenhagen/RH failed to achieve the goal with a coverage slightly below 95% – the other regions achieved the goal, although not much more than that; the national average for this patient population is 96%. In the context of patients treated in general practice none of the regions achieve the goal and the national average for this patient population is 88%. Region Zealand was doing worst with 84%, whereas the region in which I live, Region Midtjylland, was doing best with a 92% coverage. Of the diabetes units, Rigshospitalet, “one of the largest hospitals in Denmark and the most highly specialised hospital in Copenhagen”, seems to also be the worst performing hospital in Denmark in this respect, with only 84 % of patients being screened – which to me seems exceptionally bad considering that for example not a single hospital in the region in which I live is below 95%. Nationally roughly 20% of patients with micro- or macroalbuminuria are not on ACE-inhibitors/Angiotensin II receptor antagonists.

Eye examination: The main process goal here is at least one eye examination every second year for at least 90% of the patients, and a requirement that the treating physician knows the result of the eye examination. This latter requirement is important in the context of the interpretation of the results (see below). For patients treated in diabetes units, four out of five regions achieved the goal, but there were also what to me seemed like large differences across regions. In Southern Denmark, the goal was not met and only 88 % had had an eye examination within the last two years, whereas the number was 98% in Region Zealand. Region Zealand was a clear outlier here and the national average for this patient population was 91%. For patients treated in general practice no regions achieved the goal, and this variable provides a completely different picture from the previous variables in terms of the differences between patients treated in diabetes units and patients treated in general practice: In most regions, the coverage here for patients in general practice is in the single digits and the national average for this patient population is just 5 %. They note in the report that this number has decreased over the years through which this variable has been analyzed, and they don’t know why (but they’re investigating it). It seems to be a big problem that doctors are not told about the results of these examinations, which presumably makes coordination of care difficult.

The report also has numbers on how many patients have had their eyes checked within the last 4 years, rather than within the last two, and this variable makes it clear that more infrequent screening is not explaining anything in terms of the differences between the patient populations; for patients treated in general practice the numbers are still here in the single digits. They mention that data security requirements imposed on health care providers are likely the reason why the numbers are low in general practice as it seems common that the GP is not informed of the results of screenings taking place, so that the only people who gets to know about the results are the ophthalmologists doing them. A new variable recently included in the report is whether newly-diagnosed type 2 diabetics are screened for eye-damage within 12 months of receiving their diagnosis – here they have received the numbers directly from the ophthalmologists so uncertainty about information sharing doesn’t enter the picture (well, it does, but the variable doesn’t care; it just measures whether an eye screen has been performed or not) – and although the standard set is 95% (at most one in twenty should not have their eyes checked within a year of diagnosis) at the national level only half of patients actually do get an eye screen within the first year (95% CI: 46-53%) – uncertainty about the date of diagnosis makes it slightly difficult to interpret some of the specific results, but the chosen standard is not achieved anywhere and this once again underlines how diabetic eye care is one of the areas where things are not going as well as the people setting the goals would like them to. The rationale for screening people within the first year of diagnosis is of course that many type 2 patients have complications at diagnosis – “30–50 per cent of patients with newly diagnosed T2DM will already have tissue complications at diagnosis due to the prolonged period of antecedent moderate and asymptomatic hyperglycaemia.” (link).

The report does include estimates of the number of diabetics who receive eye screenings regardless of whether the treating physician knows the results or not; at the national level, according to this estimate 65% of patients have their eyes screened at least once every second year, leaving more than a third of patients in a situation where they are not screened as often as is desirable. They mention that they have had difficulties with the transfer of data and many of the specific estimates are uncertain, including two of the regional estimates, but the general level – 65% or something like that – is based on close to 10.000 patients and is assumed to be representative. Approximately 1% of Danish diabetics are blind, according to the report.

Foot examinations: Just like most of the other variables: At least 95 % of patients, at least once every second year. For diabetics treated in diabetes units, the national average is here 96% and the goal was not achieved in Copenhagen (CRD) (94%) and northern Jutland (91%). There are again remarkable differences within regions; at Helsingør Hospital only 77% were screened (95% CI: 73-82%) (a drop from 94% the year before), and at Hillerød Hospital the number was even lower, 73% (95% CI: 70-75), again a drop from the previous year where the coverage was 87%. Both these numbers are worse than the regional averages for all patients treated in general practice, even though none of the regions meet the goal. Actually I thought the year-to-year changes in the context of these two hospitals were almost as interesting as the intraregional differences because I have a hard time explaining those; how do you even set up a screening programme such that a coverage drop of more than 10 % from one year to the next is possible? To those who don’t know, diabetic feet are very expensive and do not seem to get the research attention one might from a cost-benefit perspective assume they would (link, point iii). Going back to the patients in general practice on average 81 % of these patients have a foot examination at least once every second year. The regions here vary from 79% to 84%. The worst covered patients are patients treated in general practice in the Vordingborg sygehus catchment area in the Zealand Region, where only roughly two out of three (69%, 95% CI: 62-75%) patients have regularly foot examinations.

Aside from all the specific indicators they’ve collected and reported on, the authors have also constructed a combined indicator, an ‘all-or-none’ indicator, in which they measure the proportion of patients who have not failed to get their Hba1c measured, their feet checked, their blood pressure measured, kidney function tests, etc. … They do not include in this metric the eye screening variable because of the problems associated with this variable, but this is the only process variable not included, and the variable is sort of an indicator of how many of the patients are actually getting all of the care that they’re supposed to get. As patients treated in general practice are generally less well covered than patients treated in the diabetes units at the hospitals I was interested to know how much these differences ‘added up to’ in the end. For the diabetes units, 11 % of patients failed on at least one metric (i.e. did not have their feet checked/Hba1c measured/blood pressure measured/etc.), whereas this was the case for a third of patients in general practice (67%). Summed up like that it seems to me that if you’re a Danish diabetes patient and you want to avoid having some variable neglected in your care, it matters whether you’re treated by your local GP or by the local diabetes unit and that you’re probably going to be better off receiving care from the diabetes unit.

Some descriptive statistics from the appendix (p. 95 ->):

Sex ratio: In the case of this variable, they have multiple reports on the same variable based on data derived from different databases. In the first database, including 16.442 people, 56% are male and 44% are female. In the next database (n=20635), including only type 2 diabetics, the sex ratio is more skewed; 60% are males and 40% are females. In a database including only patients in general practice (n=34359), like in the first database 56% of the diabetics are males and 44% are females. For the patient population of children and young adults included (n=2624), the sex ratio is almost equal (51% males and 49% females). The last database, Diabase, based on evaluation of eye screening and including only adults (n=32842), have 55% males and 45% females. It seems to me based on these results that the sex ratio is slightly skewed in most patient populations, with slightly more males than females having diabetes – and it seems not improbable that this is to due to a higher male prevalence of type 2 diabetes (the children/young adult database and type 2 database seem to both point in this direction – the children/young adult group mainly consists of type 1 patients as 98% of this sample is type 1. The fact that the prevalence of autoimmune disorders is in general higher in females than in males also seems to support this interpretation; to the extent that the sex ratio is skewed in favour of males you’d expect lifestyle factors to be behind this.

Next, age distribution. In the first database (n=16.442), the average and the median age is 50, the standard deviation is 16, the youngest individual is 16 and the oldest is 95. It is worth remembering in this part of the reporting that the oldest individual in the sample is not a good estimate of ‘how long a diabetic can expect to live’ – for all we know the 95 year old in the database got diagnosed at the age of 80. You need diabetes duration before you can begin to speculate about that variable. Anyway, in the next database, of type 2 patients (n=20635), the average age is 64 (median=65), the standard deviation is 12 and the oldest individual is 98. In the context of both of the databases mentioned so far some regions do better than others in terms of the oldest individual, but it also seems to me that this may just be a function of the sample size and ‘random stuff’ (95+ year olds are rare events); Northern Jutland doesn’t have a lot of patients so the oldest patient in that group is not as old as the oldest patient from Copenhagen – this is probably but what you’d expect. In the general practice database (n=34359), the average age is 68 (median=69) and the standard deviation is 11; the oldest individual there is 102. In the Diabase database (n=32842), the average age is 62 (median=64), the standard deviation is 15 and the oldest individual is 98. It’s clear from these databases that most diabetics in Denmark are type 2 diabetics (this is no surprise) and that a substantial proportion of them are at or close to retirement age.

The appendix has a bit of data on diabetes type, but I think the main thing to take away from the tables that break this variable down is that type 1 is overrepresented in the databases compared to the true prevalence – in the Diabase database for example almost half of patients are type 1 (46%), despite the fact that type 1 diabetics are estimated to make up only 10% of the total in Denmark (see e.g. this (Danish source)). I’m sure this is to a significant extent due to lack of coverage of type 2 diabetics treated in general practice.

Diabetes duration: In the first data-set including 16.442 individuals the patients have a median diabetes duration of 21,2 years. The 10% cutoff is 5,4 years, the 25% cutoff is 11,3 years, the 75% cutoff is 33,5 years, and the 90% cutoff is 44,2 years. High diabetes durations are more likely to be observed in type 1 patients as they’re in general diagnosed earlier; in the next database involving only type 2 patients (n=20635), the median duration is 12.9 years and the corresponding cutoffs are 3,8 years (10%); 7,4 years (25%); 18,6 years (75%); and 24,7 years (90%). In the database involving patients treated in general practice, the median duration is 6,8 years and the cutoffs reported for the various percentiles are 2,5 years (10%), 4,0 (25%), 11,2 (75%) and 15,6 (90%). One note not directly related to the data but which I thought might be worth adding here is that of one were to try to use these data for the purposes of estimating the risk of complications as a function of diabetes duration, it would be important to have in mind that there’s probably often a substantial amount of uncertainty associated with the diabetes duration variable because many type 2 diabetics are diagnosed after a substantial amount of time with sub-optimal glycemic control; i.e. although diabetes duration is lower in type 2 populations than in type 1 populations, I’d assume that the type 2 estimates of duration are still biased downwards compared to type 1 estimates causing some potential issues in terms of how to interpret associations found here.

Next, smoking. In the first database (n=16.442), 22% of diabetics smoke daily and another 22% are ex-smokers who have not smoked within the last 6 months. According to the resource to which you’re directed when you’re looking for data on that kind of stuff on Statistics Denmark, the percentage of daily smokers was 17% in 2013 in the general population (based on n=158.870 – this is a direct link to the data), which seems to indicate that the trend (this is a graph of the percentage of Danes smoking daily as a function of time, going back to the 70es) I commented upon (Danish link) a few years back has not reversed or slowed down much. If we go back to the appendix and look at the next source, dealing with type 2 diabetics, 19% of them are smoking daily and 35% of them are ex-smokers (again, 6 months). In the general practice database (n=34.359) 17% of patients smoke daily and 37% are ex-smokers.

BMI. Here’s one variable where type 1 and type 2 look very different. The first source deals with type 1 diabetics (n=15.967) and here the median BMI is 25.0, which is comparable to the population median (if anything it’s probably lower than the population median) – see e.g. page 63 here. Relevant percentile cutoffs are 20,8 (10%), 22,7 (25%), 28,1 (75%), and 31,3 (90%). Numbers are quite similar across regions. For the type 2 data, the first source (n=20.035) has a median BMI of 30,7 (almost equal to the 1 in 10 cutoff for type 1 diabetics), with relevant cutoffs of 24,4 (10%), 27,2 (25%), 34,9 (75%), and 39,4 (90%). According to this source, one in four type 2 diabetics in Denmark are ‘severely obese‘ and more diabetics are obese than are not. It’s worth remembering that using these numbers to implicitly estimate the risk of type 2 diabetes associated with overweight is problematic as especially some of the people in the lower end of the distribution are quite likely to have experienced weight loss post-diagnosis. For type 2 patients treated in general practice (n=15.736), the median BMI is 29,3 and cutoffs are 23,7 (10%), 26,1 (25%), 33,1 (75%), and 37,4 (90%).

Distribution of Hba1c. The descriptive statistics included also have data on the distribution of Hba1c values among some of the patients who have had this variable measured. I won’t go into the details here except to note that the differences between type 1 and type 2 patients in terms of the Hba1c values achieved are smaller than I’d perhaps expected; the median Hba1c among type 1s was estimated at 62, based on 16.442 individuals, whereas the corresponding number for type 2s was 59, based on 20.635 individuals. Curiously, a second data source finds a median Hba1c of only 48 for type 2 patients treated in general practice; the difference between this one and the type 1 median is definitely high enough to matter in terms of the risk of complications (it’s more questionable how big the effect of a jump from 59 to 62 is, especially considering measurement error and the fact that the type 1 distribution seems denser than the type 2 distribution so that there aren’t that many more exceptionally high values in the type 1 dataset), but I wonder if this actually quite impressive level of metabolic control in general practice may not be due to biased reporting, with GPs doing well in terms of diabetes management being also more likely to report to the databases; it’s worth remembering that most patients treated in general practice are still not accounted for in these data-sets.

Oral antidiabetics and insulin. In one sample of 20.635 type 2 patients, 69% took oral antidiabetics, and in another sample of 34.359 type 2 patients treated in general practice the number was 75%. 3% of type 1 diabetics in a sample of 16.442 individuals also took oral antidiabetics, which surprised me. In the first-mentioned sample of type 2 patients 69% (but not the same amount of individuals – this was not a reporting error) also took insulin, so there seems to be a substantial number of patients on both treatments. In the general practice sample included the number of patients on insulin was much lower, as only 14% of type 2 patients were on insulin – again concerns about reporting bias may play a role here, but even taking this number at face value and extrapolating out of sample you reach the conclusion that the majority of patients on insulin are probably type 2 diabetics, as only roughly one patient in 10 is type 1.

Antihypertensive treatment and treatment for hyperlipidemia: Although there as mentioned above seems to be less focus on hypertension in type 1 patients than on hypertension in type 2 patients, it’s still the case that roughly half (48%) of all patients in the type 1 sample (n=16.442) was on antihypertensive treatment. In the first type 2 sample (n=20635), 82% of patients were receiving treatment against hypertension, and this number was similar in the general practice sample (81%). The proportions of patients in treatment for hyperlipidemia are roughly similar (46% of type 1, and 79% and 73% in the two type 2 samples, respectively).

Blood pressure. The median level of systolic blood pressure among type 1 diabetics (n=16442) was 130, with the 75% cutoff intersecting the hypertension level (140) and 10% of patients having a systolic blood pressure above 151. These numbers are almost identical to the sample of type 2 patients treated in general practice, however as earlier mentioned this blood pressure level is achieved with a lower proportion of patients in treatment for hypertension. In the second sample of type 2 patients (n=20635), the numbers were slightly higher (median: 133, 75% cutoff: 144, 90% cutoff: 158). The median diastolic blood pressure was 77 in the type 1 sample, with 75 and 90% cutoffs of 82 and 89; the data in the type 2 samples are almost identical.

April 24, 2015 - Posted by | data, diabetes, medicine

No comments yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: